Final Report Summary - ISD (Intelligent Surgical Device)
Executive Summary:
The objective of the collaboration research project was the development of a mass spectrometry based tissue identification system capable of in-vivo, real time identification of tissues during surgical and diagnostic interventions. The underlying Rapid Evaporation Ionization Mass Spectrometry (REIMS) method has been developed prior to the project launch and according to the early feasibility studies it provides enough information for tissue characterization. In order to prove true intrasurgical applicability for a number of areas, a device interface, the surgical tools had to be adapted and extensively tested in real life situations.
In accordance with the objectives development of an advanced REIMS atmospheric interface for Xevo G2-S mass spectrometers (Waters) has been completed. Modifications on the benchtop instrument have been carried out in order to make that operating theatre compatible (surgical version). Three unit of this surgical versions for in vivo demonstrations and two additional benchtop G2-S for data collection have been delivered and installed. Electrosurgical handpieces with fume aspiration capabilities has been developed to support the demonstration and data collection activities. Addressing the data management needs separate software platforms for data collection and algorithm development has been developed.
Protocols for ex-vivo data collection and in-vivo applications have been completed, ethical permissions for the study in two locations have been granted. Data collection has been launched and in progress.
Results of preliminary studies in different areas of developments as well as representative data of applications in different surgical areas have been published extensively.
The main aim of the project was to develop the technology to a status where commercial applicability is no longer a distant possibility. The target was to develop the technology to status where data collected for validation is as close to the future application setting as possible for an R&D activity. The Consortium members achieved the above interim objective by completing most if not all the milestones described in the DoW, and were continuously working to ensure the long term survival of the project by proving the project’s ultimate commercial viability.
As a consequence of the first major communication round in June 2013, a number of 3rd parties indicated their willingness to cooperate with the project. Interest was recorded from both academic institutions and industrial partners. The decision was met to pursue these leads to try to create a long term strategy for the project. Towards the end of 2013 Waters UK expressed its interest in a closer cooperation which the consortium dully reported to the EU. In the discussions about preferred structure during the half of 2014 it became evident that the EU funding needs to be terminated if the long term funding of the project is to be secured through an industrial partnership. The EU was informed again and potential solutions discussed. We trust that this decision will secure the long term future of this most successful project, in which the Consortium Members agreed to carrion as defined in the original project plan irrespective of funding sources.
Project Context and Objectives:
The general aim of the proposed project is to show the commercial viability of the recently developed REIMS tissue identification method raising the technical status to a level where it could be introduced it into clinical environment obtaining sufficient amount of data to prove that a later regulatory process is worthwhile. The target of the R&D activity was to raise the technical status of the project to a level where it can grab enough global awareness to be able to secure the survival of the technology.
REIMS-based tissue identification is implemented by the on-line analysis of the aerosol formed on the electrosurgical dissection or coagulation process. MediMass Ltd. has been developing REIMS technology and corresponding instrumentation and bioinformatics toolkit for 3 years. Pre-clinical and clinical experiments revealed that the concept of real-time, in-situ, in-vivo tissue identification by REIMS technique is feasible, however standardisation and development of the device and large-scale clinical studies are needed for the accurate definition of the role of this phenotyping approach in the hospital patient journey.
The developed mass spectrometric devices will provide mass spectrometric chemical profiles of corresponding tissues, information which in itself cannot be interpreted by medical professionals. Furthermore, since the profile data does not carry “absolute” information, it can only be interpreted by comparison with a set of authentic data. Comparison will be implemented by a combination of unsupervised and supervised statistical multivariate pattern recognition algorithms. That is, instead of comparing unknown spectrum to large set of known entries, the database is converted into a histologically assigned data space, where the individual spectra represent single data points. Unknowns are identified by localizing the new datapoint on this data map, measuring its distance from surrounding data clouds (each representing a type of tissue) and classify it into the closest. This strategy allows practically real-time analysis, since the database conversion is performed prior to medical intervention. As it was already pointed out, the identification strategy calls for a large set of authentic data. One of the most important objectives of the proposed research is the construction of this database, however not only with classical histological assignment.
We planned to collect all relevant biochemical (proteomic and genetic marker statuses) and clinical (survival, drug sensitivity, metastatic behavior, aggressiveness, etc.) information and store it with the spectral information in an SQL-type database structure. This database allows the deciphering of correlations between our lipidomic/metabolomic data and status of traditional markers, and definition of novel, combinatorial biomarkers for the identification of “classic” phenotypes. Availability of clinical data and potential follow-up of patients allows the establishment of correlations between the characteristics of disease and tissue lipidome/metabolome. Well-established correlations (e.g. between abundance of certain complex lipid and tendency for metastasis formation) enable us to define novel predictive markers, and eventually a new type of (i.e lipidomic/metabolomic) clinical phenotyping system. We believe that small molecular markers give more comprehensive information on the short and mid-term status of the body than genetic or proteomic markers. Furthermore, these species also provide information on extragenomic factors such as- the composition and effect of associated microbial flora, which is a critically important factor in case of the diseases of gastrointestinal or urogenital systems.
The first step of proposed project was to establish the technical conditions for testing the methodology in clinical environment. Instrument development activities comprised the development of electrosurgical handpieces with auxiliary consumables, ion transfer devices, operating room compatible instrument chassis and so-called mass spectrometric front-end solutions. Development of necessary instrumentation is described in details in chapter “Development of dedicated mass spectrometric instrumentation for use in human medical environment”, Work package 1.
After providing the appropriate instrumental background the next step was the introduction of a mass spectrometric tissue identification device into operating theatre and optimizing the method to facilitate the translation of this new technique into standard operation practice. The clinical component of the proposal was also divided into two distinct work packages. One (Work package 2) focused on the practical utilization of the technique in surgical environment and collection of limited amount of in-vivo data (limited by ethical constraints), while the other,(Work package 3) comprised the large-scale acquisition of histologically authentic data using fresh ex-vivo samples. The in-vivo activities, grouped in Work package 2, targeted the collection of valuable feedback of surgeons from the different surgical areas regarding the special features related to the applicability of the REIMS based tissue identification method. The aim of the ex-vivo data collection tasks in Work package 3 was primarily a systematic collection of REIMS data from a range of tissue types in order to populate spectral databases for the effective training the recognition method. The larger spectral library used as a training set for a surgical area the better in-vivo classification efficacy can be achieved. While the full spectral library collection from the ex-vivo samples (3rd deliverable of WP3) provides a different aspect for understanding the fine structure of the tumours and its close proximity, utilizing the fine spatial resolution of the imaging techniques.
Work package 4 comprised development tasks targeting more universal biological tissue identification. The first task aims at a development of a spatial logging system by linking mass spectrometric data with spatial information using two different approach. The first (Simultaneous Camera Report, D4.1) targeted the exploration of the possibilities to assess spatial information by application of cameras around the area of intervention and storing the MS identification results together with the video captures in a time-synchronized manner. The other ( Spatial Logging System, D4.2) relied on instrumental background already available in some surgical environment where navigation devices are extensively used (such as neuro-navigation system in case of brain surgery). The simultaneous use of navigation with REIMS tissue identification, especially together with pre-operative imaging results, on one hand helps the surgeon to assess the spatial shift relative the pre-operative conditions and also give an indication where Mass spectrometric identification is needed (i.e. close to tumor border). The second task of this Work package (Spectral Interconversion Algorithm, D4.3) has the objective to provide an tool to convert spectral information collected with different mass spectrometers and/or different ionization methods. Such a tool could facilitate the deeper penetration of the scientific community as the currently instrument and ionization technique dependent datasets would become comparable.
The goal of the tasks collected in Work package 5 was to create a robust and stable bioinformatical toolkit for biological tissue identification and pattern analysis. Besides the improvement of the currently applied PCA/LDA based tissue classification algorithms, targeted engines for universal marker screening on the collected dataset have been planned to develop. The ultimate goal of this study is to identify new metabolomics/lipidomic fingerprints for tumor phenotyping and ultimately improve the clinical outcome in patients with malignant disease.
WP7 accumulated tasks that were targeted towards the long term viability of the R&D activity that is dissemination activities and IP maintenance. The goal was to raise sufficient awareness – both scientifically among the general public –, and to consolidate the IP portfolio such that it would form a basis for a long term product development.
Project Results:
Development of dedicated mass spectrometric instrumentation for use in human medical environment (WP1)
Objectives and introduction
The general objective of this work package was the development of instruments for Rapid Evaporative Ionization Mass Spectrometry (REIMS) based intraoperative tissue identification technology, which is suitable for use in the medical environment. We planned to develop electrosurgical intervention devices (handpieces, endoscopic snares…) that are capable to collect and transfer the surgical fume to the inlet of the mass spectrometer as well.
To provide an overview of the context of the tasks within the work package, after a short introduction of the ISD device and its accessories, the main functional units and their connection to the deliverables will be presented beforehand. Afterwards, a high level summary will be provided on the task level progress.
The device converts the chemical information of the surgical fume arriving from the area of intervention through the handpiece into mass spectra, which can be used for the characterization and identification of the tissue. The surgical fume is transferred to the atmospheric interface by a venturi airjet pump embedded in the venturi chamber attached onto the front plate of the atmospheric interface. The surgical fume is then driven to the atmospheric interface through a sampling capillary by the aspiration of the vacuum pumps of the mass spectrometer. The droplets and particles containing pre-formed ionic species then will be declustered assisted by collisions on the surface (of optimized geometry and position) and evaporation (due to low pressure and elevated temperature) resulting in ions that can be subject to separation according to their mass/charge ratio in the mass spectrometer. The acquired raw data are transferred to the database managed by Medimass/ISD SW. There a new model can be built using the histopathologically validated data sets, or in case of surgical applications, the resulting spectra can be used for real-time tissue characterization based on a model prepared earlier.
The device contains the following functional elements/units (1. Figure):
Handpiece
This is an electrosurgical intervention device, which has the functions of general electrosurgery, and is in addition to the standard functions suitable for sampling aerosol generated by the electrosurgical device for REIMS analysis. (see more in the report for deliverable 1.3)
Venturi chamber
This is the embodiment attached to the atmospheric interface equipped with a venturi airjet pump to transfer the surgical fume generated from the area of intervention through the handpiece into the atmospheric interface. (see further details in deliverable 1.1)
Atmospheric interface
This unit serves as an interface between the atmospheric pressure region of the venturi chamber and the vacuum region (of approx. 2 mbar) of the first zone of the ion optics of the mass spectrometer. Since this is the area where the preformed ionic species will be declustered from the droplets originally formed in the area of the surgical intervention, the design has a significant role in the generated ion intensity and as such, the overall sensitivity of the instrument. (see further details in deliverable 1.1 1.5)
Mass spectrometer
This is the unit that analyzes the generated ionic species according to their mass/charge ratio. For the surgical version, the original device will be redesigned to be compatible with the special requirements that originate in the application. (see further details in deliverable 1.1 1.5)
Software and database
Beside instrument control, raw data acquisition and pre-processing, the software had to answer all requirements that originate in the task of tissue characterization. Pattern-based tissue identification requires a database in which the tissue-specific spectra are stored (that are collected during the validation process), and a classification algorithm (real time recognition). The development of a software that covers all these needs and is able to handle different mass spectrometric platforms was indispensable as well. (See further details in the reports for deliverables 1.1,1.2 an all WP 2.5)
ISD prototype using existing MS platform
Operating room compatible (surgical) version development
The challenge was to convert a Waters G2S benchtop mass spectrometer such that it could be used in an operating theatre. To facilitate the transfer between the maintenance area and the theatre, the instrument had to be mobilized by proper application of handles and wheels (Figure 2.).
The cooling air circulation had to be redesigned in a way that the exhaust air could leave the chassis through a HEPA filter, so preventing the air of the operating room from being contaminated. This was accomplished by the attachment of a filter box to the chassis containing an exhaust fan and a series of filters and appropriate pressure sensors to monitor their performance as shown in 3. Figure and 4.
Atmospheric interface development
Cold collision surface
As a first step, we explored the effect of the collision surface geometry and position on the signal intensity. After a series of experiments with different shapes sizes of the collision surface, we concluded that the spherical collision surface with a given diameter positioned at a specific distance from the sampling capillary provided the optimal ion intensity. In order to do away with potential memory effectsdue to depositing on the collision surfaceheating was introduced to the collision surface. (5. Figure)
Heated collision surface
It was soon recognized that for the elimination of the depositing, a relatively high surface temperature is needed. The maintenance of this high temperature at normal operating conditions (continuous gas stream from the sampling capillary) requires quite a high heating power concentrated on in a relatively small volume. As the conclusion of a series of experiments, we settled with the application of a coil made of Kanthal strip, operating as a resistive heater at a high surface temperature at normal REIMS operating conditions (6. Figure).
Apart from preventing deposit from building up, the high collision surface temperature proved to be advantageous for the declustering process as well. A magnitude of intensity gain could be achieved in the REIMS spectra by heating the collision surface (7. Figure).
In conclusion an atmospheric interface for the Waters Xevo G2S mass spectrometers that provides intensive and stable signals for tissue recognition has been successfully developed.
The instrument has been converted into a device that is compatible with a surgical environment. Three of the surgical version instruments have already been installed and are being currently tested in a surgical environment. Apart from these devices, two original bench top Xevo G2S instruments will be used for data collection.
Software development
The task related to the software background can be divided into three main groups.
Instrument control
As the aim was to develop an instrument that does not require special expertise for use and maintenance, we decided that the instrument control functions have to be handled by the ISD software. Thus most of the tune parameters (to set the elements of the ion optics) will not be available for the end users but will be set by the system administrator for the optimal value. With this apart from the ease of use, we can assure that the acquired data will acquired with the same tune settings. To achieve this aim an interface program layer has been developed in close cooperation with the manufacturer.
workflow related development
Based on the feedback on the previous studies several suggestions and ideas emerged to improve the safety of the data collection and the accuracy of the histopathological classification by more strict sampling protocol. Implementation of the changes in the data collection protocol into the software has been decided. A user interface that follows the protocol step by step has been planned and its development has been launched. Additional features like handling and storing images of the samples before and after REIMS sampling and a user interface for assignment of the actual REIMS sampling point and the histopathology of the surrounding tissues will be implemented as well.
classification algorithm and data processing
As the currently applied mass spectrometric platform provide higher resolution data having quite different characteristic that the data structure we used before, significant changes had to be, and still have to be implemented to our data handling toolkit. Besides the benefits of the high resolution data collection (like separation of analytes with the same nominal molecular weight but different elemental composition, which provide a deeper view into the fine structure of the lipidomic fingerprint) it is more sensible to external parameters, like temperature and its effect on the performance of the elements of the electronics. To handle this sensitivity, a continuous adjustment of the instrument calibration using a selected molecular peak (Lock-mass) is required to maintain the resolution through the measurements. Apart from this feature we identified tasks around the noise filtering as, according to our experiences so far, we are confident that some of the periodic noise we experienced only with this mass spectrometric platform can be separated from the signals containing the chemical information, thus making the tissue specific differences more visible. To support the development of such a noise filtering algorithms development of a separate software platform (Offline Data Mining (ODM)) has been launched. Further results and details on these developments can be found in the chapter about the last task in this work package (Finetuning the ISD prototype, further step towards a medical device) or in the deliverable report (D1.5)
REIMS imager device for ex-vivo sampling
The objective of the whole project is the development of real-time tissue identification method that can be applied during surgical and diagnostic intervention. The tissue identification is based on the comparison of unknown data collected real-time to a database of authentic REIMS data by utilizing multivariate statistical pattern recognition algorithms. Therefore the applicability and performance of the tissue characterization is highly dependent on the amount of the histologically validated data points in the different application areas stored in the database. In this section the summary of the development of an automatized device, capable to high-throughput data collection, will be presented.
REIMS compatible imager platforms have been developed in two phases via the conversion of commercially available imaging stages. The first version, which meant the modification of the Prosolia DESI stage specifically purchased for this purpose by MediMass, was capable of collecting data from samples from even surfaces like tissue sections. This has been tested and compared to DESI imaging. Based on the initial experiences, an advanced imaging stage has been developed via the modification of Prosolia Flowprobe configuration (8. Figure), which provides better conditions for data collection from larger tissue specimens (which usually have uneven surfaces) since the probe positioning along the z axis is continuously adjusted according to a laser distance sensing (purchased separately at Imperial College).
For the REIMS data collection the original Flowprobe sampling head has been replaced with an electrosurgical blade holder (9. Figure). This REIMS sampling head is equipped with an aspiration channel for aerosol sampling and a replaceable blade holder (connector for tubular probe needle) and is designed to facilitate easy maintenance and decontamination. The standard tubular probe needle is similar to the needle electrode that is used during standard electrosurgical interventions.
Due to the various safety features incorporated to protect the patient, the commercial electrosurgical supply does not easily lend itself to an automated analysis platform, thus the use of a conventional power supply simplifies the experimental setup. The variations of the spectral outcome due to the application of different power supplies have been tested and found negligible.
As a result of this task a robust REIMS imager with excellent precision has been developed. For data collection a mixed application of manual and automated methodology has been considered so far. While the manual data collection has a limited throughput and also provide some more possibilities for human dependent variability, the high-throughput automated approach may contain technical features that may result in quite another outcome than the real life application. Therefore, for a proper benefit analysis a comparative study has to be executed on a larger data set.
REIMS compatible interventional devices
As electrosurgical methods are widely used during surgical interventions in different areas, the variety of the available interventional devices (handpieces) reflects the diverse needs of the different applications. Supporting the introduction of the REIMS base tissue identification method into clinical practice, universally applied handpieces were selected for the development.
Monopolar handpiece
Since the monopolar electrosurgery is one of the most generally used techniques in surgery, the monopolar handpiece with cut and coagulation functions was first selected to modify for surgical fume sampling. Based on the surgeons feedback we decided to supply four exchangeable electrodes for the different applications a needle and a blade of two different lengths of both. For the development and production of the sampling handpieces, the collaboration with Medres Ltd. has been continued. They have considerable experience in the field of medical device development. At present, development PHASE I has been closed, project documents, manufacturing drawings and instructions, material specifications and preliminary instruction for use have been finalized.
As a result of the progress handpieces for the ex-vivo manual data collection and demonstrations can be provided in sufficient quantity, and sterile handpieces in appropriate packaging will be available soon for the in-vivo demonstrations and validation tests (10. Figure and 11. Figure).
Bipolar forceps
Bipolar electrosurgery is a commonly used method of surgery as well. The device introduced to the intervention area consists of two electrodes separated by insulation. The electrical current only passes through the tissue between the two electrodes. The advantage of the bipolar technology is that the distance between the electrodes of the device (forceps stems) is only a few millimetres, therefore only very small sections of tissue are involved in the electric circuit. As opposed to the monopolar devices, bipolar forcepses are not applied for cutting, but much more for coagulations, since these use lower current density due to the much smaller tissue resistance. Bipolar forcepses are frequently used in neurosurgery, as they are precise, less invasive and stay current-free.
After a thorough search among the available bipolar handpieces, a type of forceps was found that seemed to be suitable for our application. This irrigation forceps has a built-in tube and an opening near its end (12. Figure), which is normally used for dispensing irrigation liquids to the surgical site. We have found that this type of forceps is applicable for aerosol sampling using the irrigation channel for the surgical fume sampling. As the amount of the generated aerosol is much smaller, due to the less invasive nature of this technique, the fume transfer tube is connected directly to the sampling capillary of the mass spectrometer and the aerosol is aspirated by the vacuum pumps of the mass spectrometer.
Sampling probe
During the consultations with the surgeons the need emerged for a sampling probe that is especially designed for data collection only without features supporting intervention processes like cutting. These devices could be effectively used to analyse the residual tissue layers on the wall of the cavity after the removal of tumours to ensure that the margin is clear. As this probing concept could be effectively utilized as a diagnostic approach, feasibility studies have been launched to explore its potential benefits.
To summarize the results in this area, we can conclude that the development steps required for the mass production of a disposable monopolar handpiece have been dully completed. Manufacturing phase IIa has been closed, so a critical milestone towards the transferable design has been achieved. The required amount for the data collection activity is available. The sterile end products will be available soon for the in vivo demonstrations. The bipolar applications can be supported with commercially available intervention devices. Preparations for probe development have been launched.
Adaptation to endoscopy
The main objective of the endoscopy sub-project was to couple the REIMS technology with endoscopy and produce a device capable of the in-vivo testing of gastrointestinal lesions in course of diagnostic interventions.
Endoscopic snare
The snare device comprises a hand tool, which is used for the opening and closing of the snare (by pushing the loop out of the endoscopic device and pulling it back on) and a long stainless steel thread which ends in the snare loop itself. In order to isolate the snare from the body of the endoscopic device, the stainless steel thread and the loop are surrounded by a plastic tube housing. Since the outer diameter of the plastic (HDPE or PTFE) housing is slightly smaller (<0.2 mm difference) than the internal diameter of the working channel of the endoscope, the electrosurgical aerosol containing charged particles needs to be aspirated through the housing, as an ion transfer device. There are number of endoscopic devices available with multiple work channels, however aspiration through an alternative channel was proven to be less efficient due to the inherent distance between channels and also due to the interference with the optical elements (camera, light source). Employing the snare housing as an ion transfer device has raised a number of problems including the blocking of the housing by the tissue being dissected, as it is shown on 13. Figure .
When a polyp is captured by the endoscopist, its basal part is pulled against the housing before the actual dissection takes place. As a result, the aerosols formed on dissection cannot be aspirated through the housing, since its entrance is completely blocked. Further problem with the snare housing was associated with its connection to the Venturi air jet pump-based aerosol aspiration system, since the housing tube has a dead end on the outer (handpiece) end of the device. The potential neutralization of charged particles on the bare metallic surface of the stainless steel thread was perceived as an additional problem. In order to overcome the second problem (i.e. connecting the snare housing to the aerosol aspiration system of the mass spectrometer) and also to minimize the extent of modifications made to commercially available, fully approved snares, a snare equipped with irrigation capability (manufacturer) was used as an initial experimental setup. Since the irrigation function requires the flushing of the snare with physiological saline, these snares feature a Luer port on the outer end of the device, which can easily be connected to the existing aerosol aspiration system. Due to the potential risk of short-circuiting the snare with the flushing liquid, the snare surface is covered by a polymeric varnish film in this device, which gives an efficient solution for the last problem mentioned above (neutralization of charged particles). The problem of blocking the housing inlet with the sample itself could not be solved without modifying the device. In this case the wall of the housing tubing was fenestrated close to its distal end. The shape, pattern and number of fenestrations was carefully optimized to make a compromise between optimal aerosol sampling efficiency and the mechanical stability of the plastic tube housing. The resulting slightly modified endoscopic snare was connected to the modified Xevo G2-S surgical mass spectrometer system and proof-of-concept data was collected using food-grade post-mortem porcine colon and gastric mucosa as samples.
Special accessories for endoscopic applications
Since the endoscopic application of the electrosurgical intervention occurs in cavities of the human body, the applied fume aspiration from these areas has some special features. To prepare for those the following additional hardware elements have been developed.
Triggerbox
As the intervention area has a finite volume, and the air resupply through the work channel of the probe is also limited, the continuous aspiration of the gas should be avoided in most of the cases. To address this - although it was not mentioned among the tasks of the grant - we have developed a device which monitors the operation of the electrosurgical handpiece, and depending on this, is capable to provide trigger signals towards external devices like the bypass valve attached to the exhaust of the venturi chamber, the gas supply of the venturi air jet pump and the mass spectrometer. With this synchronization, only for the duration of the actual intervention, both the fume aspiration and data acquisition be switched on. To provide flexibility for the users, the triggerbox can operate in continuous mode as well, then all the triggers are switched on independently on the status of the handpiece.
Liquid separator
Surgical handpieces equipped with aerosol aspiration systems tends to aspirate not only aerosol but also liquid accumulated in the surgical area. Aspirated liquid interferes with ion transfer and can seriously damage the mass spectrometer, so to handle this problem a route scouting of the development possibilities of a liquid separation device has been carried out. Several approaches have been considered, prototypes (1. optical droplet sensor drived valve system, 2. continuous centrifugal separator, 3. absorption based separator and a 4. conventional liquid trap) have been manufactured and tested. Two of the solutions (1 and 2) can be used in continuous mode, which is a significant benefit as they do not require additional care to change the buffer. At this stage of development we can say, that the most innovative technical solutions (1-3) are either too complex to provide a proper base for a development of a disposable or sterilizable accessory or still suffer from compromised operational reliability (1). Therefore we have to iterate further to find more simple but effective designs capable to operate in continuous mode. On the meantime we use the conventional liquid trap and special attention will be paid to emptying the buffer in time.
In conclusion, the REIMS technology was successfully coupled with both gastroscopy and colonoscopy devices. The resulting iEndoscope setup features mostly commercially available, fully approved medical devices, at least regarding the parts used in the close proximity of the patient. This fact guarantees that the instrument can enter into formal data collection without any additional regulatory approval procedure. The special requirements of the endoscopic environment has been properly addressed by development of liquid separator and triggerbox device.
Finetuning the ISD prototype, further step towards a medical device
This task was defined to collect all activities around fine-tuning the system pre-prototype obtained according to deliverable 1.1. The developments were primarily driven by the user feedback from surgeons, pathologists and research fellows, but some of the new features are just the natural implementation of the new ideas targeting the organized improvement of the system and its SW background.
Hardware modifications, atmospheric interface
Isolator plate
During the risk assessment procedure, minor modification requests emerged with respect to the atmospheric interface of an operating theatre compatible mass spectrometer under the following hypothetic fault condition. The surgical fume transfer tube gets filled with conductive material (saline) and thus creates galvanic contact between the patient and the interface chamber. As the interface chamber is connected to the technical earth, while the patient is connected to the patient earth, due to the possible potential differences between the two earth circuits, current might be generated with all its disadvantageous consequences.
To mitigate this risk, the interface chamber has been separated from the mass spectrometer with an isolating plate, and a patient - earth connection has been installed onto the interface chamber Therefore, under the fault condition described above, the separate earth circuits remain unconnected.
Integrated power supply and temperature control of the heated coil
For the simple and convenient setting of the atmospheric interface parameters (coil offset voltage and coil heating current), the external controlled current power supplies have been replaced with an alternative solution involving the available supply channels on the mass spectrometer. Besides producing simpler thus safer equipment, the coil offset setting from this point could be adjusted from the mass spectrometric tune page, and also could be logged to the experimental file in the acquisition folder.
In order to find a truly robust solution is to close the control loop with a sensor capable of capturing temperature on the kanthal coil surface and modulating the supply current according to the measured values, investigations have been carried out to find a proper method for temperature sensing. As a result of this development cycle, a prototype of a closed loop coil heating control circuit has been built based on the temperature sensing by a remote IR sensor (14. Figure).
So at this point, we can state that we have a backup solution at hand for the coil temperature control, although the integration to the mass spectrometer is still ahead. In the event of the controlled voltage power supply functioning adequately witnessed by the different users at various locations, that version will be used. In order to provide a backup solution, however, we will also continue our experiments on the closed loop temperature control approach
Software development
During the early data collection campaigns, a quite a number of important experiences was collected, which initialized the development of a new data collection module for the ISD Software package. The aim of this development is to improve the quality of the data collection by decreasing the possibilities of the human- and workflow-related errors by the application of barcode labeling and handling, and to provide further interfaces for more accurate referencing between the tissue images and the mass spectrometric data acquisition. New features have been introduced as well to support the more detailed histopathological validation and its better organized administration.
Along with the increasing amount of collected data from the different areas, a need has arisen for a support software platform, where new algorithms for the different phases of the data processing (noise filtering, background removal, feature selection, etc) could be conveniently tested and evaluated without disturbing interference with the data collection. For the demonstration of the role and performance of this offline data mining platform (ODM), please see the results of an algorithm for the selective removal of a periodic chemical noise from the mass spectra on 15. Figure.
Handpieces
To summarize the results in this area we can conclude that the mass production of a disposable monopolar handpiece has been organized. Manufacturing development phase IIa has been closed, biocompatibility tests have been specified. For the sterile product development the dose map protocol has been finalized. The validation tests were always planned to be outside the scope of the EU funded project, and will be completed during the latter part of 2014.
The required amount for the data collection activity is available. The sterile end products will be available soon for the in vivo demonstrations. The bipolar applications can be supported with commercially available intervention devices. Preparations for probe development have been launched.
As a summary, it can be stated the ISD system with all of its components become nicer safer, more integrated. The software background stepped forward towards a higher quality and more marketable product, and with the support of the ODM platform we have an effective toolkit to utilize further the tremendous amount of valuable array of information that was hidden in the forest of the mass spectral fingerprints.
Summary for WP1
To summarize the progress of the tasks belonging to this work package (WP1) all the elements of the technical background has been developed and the system is ready to support the data collection activities grouped in the other work packages. Development related to mass spectrometers including atmospheric interface and the area of operating room compatibility resulted quite matured solutions through several engineering iterations. We managed to either develop or identify commercially available interventional devices to support data collection in the targeted surgical areas. Special requirements related to Endoscopic applications have been properly answered. REIMS imager device to support automated data collection has been developed. Software development targeting the integration of the waters mass spectrometric platform and data processing protocols specially required for the high resolution nature of the applied mass spectrometric platform has been launched
Clinical development and in-vivo data collection (WP2)
Objective
The aim of this work package is the introduction of a mass spectrometric tissue identification device into operating theatre and based on the early user feedbacks optimize the method to facilitate the translation of this new technique into standard operation practice.
In-vivo data collection was planned to start in 2014. The inclusion of hospital units was always an issue for the project management (see the discussion in section 3.2.2) and due to the potential early termination the answer to this question was delayed. As a result Asklepios Medical School (AMS) played only an advisory role in the project (the doctors and R&D personnel have not been remunerated for their services), no formal link with the healing unit has been established, most of the work therefore has been allocated to institutions belonging to Imperial College. This resulted in that some of the areas, where the experiments were allocated to AMS (like brain surgery), got a bit behind the schedule due to the delays caused by the uncertainty whether the grant funding will be terminated early.
Optimal settings and in vivo protocols for surgical interventions
This task comprises activities to generate the necessary documents to facilitate the safe and reliable experimental work in the targeted surgical area including documents providing clear user instructions for the system applied and protocols to ensure the quality of the sample collection and handling.
Operation instructions for the available instruments
Two detailed user manual has been compiled for the two different ISD prototype containing all the necessary information to set up the specified operational parameters. One for the instrument based on Tofwerk CTOF mass spectrometer, and another for the waters Xevo-G2S based instruments. These documents can be found among the deliverables for this task (D2_1TofWerk optimal settings handling protocol, D2_1Xevo G2-S optimal settings handling protocol)
Protocols for in vivo operations
As there are significant differences in the practices and environment in the different surgical areas, besides a general protocol several other one had to be composed taken into account the special needs of the appropriate area. Apart from the liver brain and colorectal protocols most recently another one has been added to the collection to cover the breast surgery as well. Ethical approvals to carry out the protocols has been received for both locations.
Intraoperative spectra
The systematic collection of human colorectal, liver and breast tissue has been started in the past year and a half, while everything is ready for the kick-off of the collection of brain tissue. This latter is the one which is relatively , but mildlybehind the schedule. Theatre compatible surgical devices have been created and shipped to Imperial College for in the theatre use. These instruments have been used during liver, colorectal and brain surgery, and will be used in the near future for breast surgery also. Devices for Asklepios Hospitals are ready and will be shipped and put in use as soon the termination of the grant funding is finalised and a new structure is put in place. All planned units have received the necessary ethical backing for starting the procedure.
Breast
Breast cancer is the most common cancer affecting females in the United Kingdom and around the EU. In 2011, 41,523 women in the UK were diagnosed with invasive breast cancer the vast majority of whom received surgery with curative intent. Positive tumour resection margin (tumour at the inked margin) following attempted breast conserving surgery (BCS) is one of the most important contributing risk factors for ipsilateral local recurrence and it is recommended that in cases where margins are positive or “close” that further surgery is carried out to confirm that margins are clear of disease. A recent UK National Audit of screen detected breast cancers indicates that ~24% of women undergoing BCS require additional surgery to treat positive resection margins.
Our aim was to investigate whether near real-time tissue identification using REIMS is feasible for identification of normal, benign and malignant breast tissues (using formal histological examination as “gold standard”), and whether accurate predictions regarding margin status can be made based on REIMS data acquired in-vivo.
Our data suggests that the metabolic spectra of aerosols obtained from ex-vivo healthy and malignant breast tissues are sufficiently distinct to warrant more detailed investigation toward an in-vivo tool for margin control. All ethical permissions and clearance with the hospitals for in-vivo work were acquired, a theatre compatible, packaged device has been moved to Charing Cross Hospital and arrangement was made with breast consultants. We are continuing the ex-vivo breast tissue collection and the in-vivo work will launch n Q3 2014 in Charing Cross Hospital. We are also submitting a proposal for CRUK CTAAC in order to continue the work and answer the stage II questions.
Colorectal
Colorectal cancers are still one of the leading morbidities in US and Europe, while gastric cancers have very poor outcomes, thus an early diagnosis is an essential factor for long term survival. Rapid Evaporative Ionization Mass Spectrometry (REIMS) is a technique developed for the in-vivo characterization of human tissues by mass spectrometric analysis of the aerosol released during electrosurgical dissection. The ionization technique can also be combined with an endoscopic snare, allowing an in-situ, in-vivo analysis within the gastrointestinal lumen. The aim of this study was to develop and characterize endoscopic and colonoscopic setups, and to test the method and classification performance in-vivo. An ex-vivo tissue specific, histologically assigned spectra database is necessary for in-vivo studies, thus the first aim was to create a spectral database from ex-vivo fresh samples.
Both the monopolar and iEndoscopic setup was tested in-vivo during surgical resection of colorectal cancer or endoscopic resection of polyp in the colon. Both setups were suitable for in-vivo operation, thus the study will continue with the in-vivo phase.
Our results have shown, that there is a clear separation between healthy, cancerous and adenomatous polyp tissue spectra, as there is a clear separation between the different layers of the gastrointestinal tract. The results also suggest, that ex-vivo colorectal database is suitable for the identification of tissue acquired with our novel iEndoscopic setup. In-vivo experiments have been started, and will continue in the future for colorectal tissue and endoscopic interventions in St Mary’s Hospital and Royal Marsden Hospital.
Liver
Liver cancer is still a leading problem around the EU. Most of the patients have metastasis from other primary cancers, mainly from the gastrointestinal tract and lungs. Besides metastasis, primary liver cancers are also hard to detect, as by the time symptoms arise, usually a greater portion of the liver is affected. It has become more and more wide spread not to remove a full lobe in most cases, but to keep healthy liver parenchyma. In these cases our iKnife could be a very useful tool to detect the margins of the cancer or to identify small, unknown spots occurring in these patients throughout the liver parenchyma.
The instrument has been used in-vivo during surgical resections of liver cancer a number of times. Liver parenchyma contains a mass of blood, which could affect our results. The blood in the system does make the acquired spectra more noisy, however it did not affect the correct identification rate. In-vivo iKnife measurements were carried out in Hammersmith Hospital. In all cases, healthy liver parenchyma was sampled in the normal procedure while removing the tumour, and cancerous tissue was only sampled ex-vivo immediately after removing the tissue from the patient. All spectra were classified using the previously built ex-vivo tissue database. Our results showed >90% correct classification rate, however as healthy tissue was not removed from the patient and could not be validated by histopathologist, precise specificity and sensitivity data could not be calculated.
Brain
All brain experiments should started in Asklepios Clinic, Hamburg. The global ethical permission, including brain surgery, has been accepted now at Imperial College London, an instrument has been transported to the 14th floor surgical block of Charing Cross Hospital, and the in-vivo measurements will start around the last week of July. Our previous results collected in the Surgical Department of University of Debrecen both in-vivo and ex-vivo suggest, that iKnife coupled with bipolar forceps is capable of differentiating healthy, malignant and benign tissue both ex-vivo and in-vivo. Tissue was collected from various type of primary and metastatic cancers, benign and malignant tissue in a total number of 64 patients. Healthy tissue was collected from 6 patients, all during the surgical procedure in-vivo. The correct classification rate after leave one patient out cross validation was between 80-100%, however the number of healthy (n=30) and non-healthy (n=370) spectra has an order of magnitude difference, which biases all the results. In the future, more healthy and cancerous tissue spectra has to be acquired during surgery and immediately after surgery, still in the theatre.
Gastrointestinal surgical device
The main objective of this sub-project was to utilize the technical background developed in the first work package (see report for D1.4) for in-vivo testing of gastrointestinal lesions in course of diagnostic interventions. According to the schedule of the data collection activities, the endoscopic and laparoscopic results and in vivo demonstrations were going to start as of month 18. As we have accomplished all the technical development tasks, this part of the project is progressing in a timely manner. The produced data presented here and in the deliverable report all derive from ex vivo procedures; we are now in the position to launch the in vivo demonstrations and the data collection campaign in this field as well.
The final system was tested ex-vivo for human samples. Gastric cancer and healthy gastric mucosa samples were collected at the Department of Histopathology, St. Mary’s Hospital, Paddington, from complete gastrectomy specimens by histopathology consultants. Healthy gastric mucosa and gastric adenocarcinoma – similarly to earlier observations – were found to show markedly different spectral profiles as it is demonstrated by the multivariate statistical plot on 16. Figure.
In conclusion, the REIMS endoscopy device is ready for human testing, both from instrumental and ethical point of view. The experiments will start in July at St. Mary’s Hospital (ICL).
Summary for WP2
Most of the surgical area we originally planned to introduce this new methodology has been successfully involved into the cooperation. Although in some cases, especially in the area of brain surgery (due to abovementioned reasons), the expected numbers of case has not been achieved by the proposed time, we are confident that having passed the warm-up period the progress will be a bit accelerated.
High throughput data collection for the development of spectral database (WP3)
Objective
The aim of this work package was to establish the infrastructure and methodology for large scale acquisition of histologically assigned REIMS data set. In order to ensure the quality of the collected data compilation of a protocol for histopathological and mass spectrometric characterization of ex vivo human tissue samples was also indispensable.
Protocols for sample collection
A protocol has been compiled with all the details that is required to eliminate risks that may result diversification of the resulted data due to inappropriate sample handling. The workflow regulates also the sampling geometry to facilitate the safe histological assignation of the REIMS sampling spots. During the early data collection campaigns, a quite a number of important experiences was collected, which initialized the development of a new data collection module for the ISD Software package. The aim of this development is to improve the quality of the data collection by decreasing the possibilities of the human- and workflow-related errors by the application of barcode labeling and handling, and to provide further interfaces for more accurate referencing between the tissue images and the mass spectrometric data acquisition. New features have been introduced as well to support the more detailed histopathological validation and its better organized administration.
REIMS spectral library
As our attempt to launch the collection at Asklepios Medical School were delayed, we started all the tasks related to data collection to institutions belonging to Imperial College. The systematic collection of human colorectal, liver and breast tissue has been going on in the past year and a half at Imperial College, while everything is ready for the kick-off of the collection of brain tissues. Due to the difficulties around the data collection sites mentioned above, we did not manage to achieve the number of data entries planned for the 18th months, but we are still confident that after the successful full start of the data collection activity we would have sped up and proceeded in a timely manner. The number of samples collected is increasing each month, as more consultants are becoming involved, and new research nurses and clinical fellows are being hired.
Full spectral library
The task was aimed at the histological mapping of REIMS-accessible biochemistry with the rationale of validating the already described biomarkers (single or combinatorial) and systematic discovery of novel ones. At the first stage of the sub-project DESi-MSI was selected as the optimal imaging technology for the purpose. Since the chemical application range of DESI and REIMS show excellent overlap (both techniques detect predominantly complex lipid species), DESI does not suffer from experiment-related spectral artefacts and the technique provides sufficient sensitivity, the task was accomplished using this approach. Close to a thousand individual tumour specimens were analysed and over 2 million histologically assigned spectra were acquired. This unique database has revealed significant biochemical differences between tumours and surrounding healthy tissues and also clearly indicated that the tumour-specific biochemical alterations cross histological borders resulting in a well-defined tumour environmental biochemical segment. The observations pointed the attention to certain molecular species and their combinations and also gave evidence for the feasibility of proximal tumour detection by REIMS.
Summary for WP3
All the essential prerequisite infrastructure and methodology has been established for the large scale data collection. Though the number of collected data entries defined in the second milestone for the 18th months has not yet been achieved, we are still confident that after the successful full start of the data collection activity we would have sped up and proceeded in a timely manner. Regarding the full spectral library collection and its utilization to discover the fine structure of the tumors and their proximity we managed to achieve significant process ahead of the schedule.
Bioinformatics- spatial logging system and platform free development (WP4)
objective
In this work package tasks targeting the establishment of a platform for universal biological tissue identification has been collected. Development of a spectral interconversion algorithm to create possibility of universal application of mass spectrometric datasets generated with different ionization methods (laser, bipolar and unipolar electrosurgical devices) has been scheduled. A spatial logging system to correlate spatial and spectral data has been planned to be developed as well.
Simultaneous camera report
The object of the data recording is to correlate the mass spectrum with the exact location in the tissue where it was generated from. In some cases – for example during most brain surgeries – there is a spatial logging system or neuro-navigation which helps the surgeon with precise localization in the brain. However, in most abdominal and thoracic cases, there is no such navigation system available, the surgeon can only relies on his/her own eye-sight. Consequently, in these cases we need a simple localization system to validate our in-vivo data.
Originally, the implementation of the cameras was designed to occur simultaneously with the REIMS ISD data collection. Considering the fact, however, that the application of the ISD evokes quite a great deal of stress for the majority of the volunteering surgeons and their teams, we decided not to intensify the already existing pressure by involving new devices, such as cameras. Furthermore, the ethical issues arising from the process of embedding our methods in the local environment can also be handled more easily, if the modifications affecting the usual procedures are implemented in a gradual way. As a result, launching our validation experiments is deferred to a later date.
Proof of concept experiments has been performed to outline the possibilities to achieve the original objectives in this area. Feedbacks collected during these early studies convinced us that further efforts to develop a more sophisticated system should be deferred to a later phase of the project when the environment for in vivo validation studies becomes well established (expected by end of 2014
Spatial logging system
This sub project is aimed at the simultaneous use of navigation with REIMS tissue identification This combination on one hand helps the surgeon to assess the spatial shift relative the pre-operative conditions and also give an indication where Mass spectrometric identification is needed (i.e. close to tumor border).
We originally planned to carry out experiments targeting the integration of neuronavigation devices with the ISD technology at Asklepios Medical School. As our attempt to launch this collaboration failed, we had to reorganize the data collection activities to institutions belonging to Imperial College.
Due to difficulties mentioned above, this sub project is still in the preparation phase, thus a bit behind the schedule. Once we manage to launch the collaboration with the Neurosurgical Department of Charing Cross Hospital (Imperial College), which is planned in second half of 2014, the progress in this area will be more visible.
Spectral interconversion algorithm
Data fusion strategy to convert raw mass spectral intensities of biological samples measured by distinct desorption ionization methods and/or instrumental setups to cross-platform normalized analyte profiles has been developed. This strategy starts with the database driven analyte peak annotation and summarization, and after that, it derives analyte-specific normalization factors to adjust intensities of a common subset of analytes to a comparable scale between platforms. It is assumed that distinct MS-based platforms capture a similar set of chemical species in a biological sample, though these species may exhibit platform-specific molecular ion pattern distributions (17. Figure). The method was validated on a dataset obtained by laser desorption ionization (LDI), Desorption Electrospray Ionization (DESI) and Rapid Evaporative Ionization mass spectrometric (REIMS) analysis of porcine tissues. We demonstrate the capacity of our method to reduce MS-platform specific variation resulting in 1) high inter-platform concordance coefficients of analyte intensities 2) clear principal component-based clustering of analyte profiles according to histological tissue types, irrespective of the used desorption ionization technique or mass spectrometer and 3) accurate “blind” classification of histological tissue types using cross-platform normalized analyte profiles.
MS based tissue profiling and imaging techniques are beginning to translate into the clinical domain. In order to provide accurate chemical tissue identification, large histologically assigned databases, likely to be acquired over many years on multiple instruments and platforms, are required. The proposed XMS data fusion method offers a promising approach for converting raw mass spectral intensities of biological samples measured by distinct desorption ionization methods and/or instrumental setups to cross-platform normalized analyte profiles, We have demonstrated the capacity of the XMS method in order to sufficiently reduce platform/instrument specific variation in MS data resulting in high inter-platform concordance coefficients of analyte intensities, accurate classification of histological tissue types, measured on multiple platforms, irrespective of the desorption ionization technique or mass spectrometer used. The proposed validation strategy can also be used for further refinement of cross-platform normalization methods. The accurate tissue identification obtained on XMS normalized data, allows imaging-based desorption ionization methods to be used for the population of large scale MS-based tissue databases which are suitable for near real time intra-operative tissue identification using REIMS technology.
Summary for WP4
Proof of concept experiments has been performed to outline the possibilities for harvesting spatial information along with the mass spectrometric data. Feedbacks collected during these early studies convinced us that further efforts to develop a more sophisticated system should be deferred to a later phase of the project when the environment for in vivo validation studies becomes well established (expected by end of 2014). A functional toolkit has been successfully developed for interconversion of spectra collected with different ionization methods.
Bioinformatics- pattern and phenotype analysis (WP5)
Objective
The goal of this project is to create a robust and stable bioinformatical toolkit for biological tissue identification and pattern analysis. Targeted engines for universal marker screening on the collected dataset have been planned to develop. The ultimate goal of this study is to identify new metabolomics/lipidomic fingerprints for tumor phenotyping and ultimately improve the clinical outcome in patients with malignant disease.
Algorithm for marker identification
The objective of this task was to define different algorithms for the search of specific tumor markers in the human database. The algorithm we used so far is based on the complete spectrum as a fingerprint, however, in case of mixed spectra (part healthy, part cancerous tissue), monitoring specific molecule peaks could help in the precise identification of the sample. We have to point this out, even if there is only a small amount of cancerous tissue in the spectra. We have tried a couple of different algorithms: First, we tried to look at the loading plots of the PCA, however the most significant peaks were the most intensive peaks, thus we moved on to more precise algorithms among which the following three were selected for the systematic screen.
Biomarker m/z feature recovery with Bayesian analysis
An empirical Bayesian analysis of variance (ANOVA) was employed in order to determine the relevance of a given molecular ion species with regards to observed tissue type discrimination. This analysis is specifically suited for high-throughput biological datasets, as it borrows the information across metabolites for improved variance estimate of the null hypotheses of non-differentially abundant chemical species.
Group deviation filtering
As a first step we search for peaks on the m/z axis in the spectra which can differentiate between the various groups. These peaks will be assigned biomarkers by a probability. The perfect biomarker would be a peak with probability 1.
Kolmogorov-Smirnov nonparametric statistical test
This algorithm is used when the distribution of the data is unknown. It can be used on any kind of data and answers the question whether two populations belong to the same distribution or not.
systematic marker screening
The object of this task was to define different algorithms for the search of specific tumor markers in the human database. We have used the previously described algorithms on different spectra acquired from healthy and cancerous human tissue. The mass of the current human database was acquired with Thermo LCQ Deca or LTQ Velos mass spectrometers. These instruments are of low resolution, thus the dataset generated does not contain individual phospholipid peaks but the sum of 2-3 different species in one peak. The high resolution Waters Xevo G2-S instruments – more suitable to provide decent data for marker screening - were placed in the theatres the past month or will be placed there early next year, thus a large amount of data collection will start from January, 2014 as part of WP2 and WP3. As soon as at least 100 healthy and 100 cancerous human spectra of the same area is acquired, we will run these algorithms again on that dataset. We have tried marker screening on a small amount of high resolution spectra acquired with the instrument Thermo Exactive, and as a large amount of data is required for the algorithms to work properly, we also tried them on low resolution LTQ data.
When comparing the results, the algorithm based on simple ANOVA and the Kolmogorov-Smirnov test pointed out more potential “biomarker” peaks, than the algorithm based on the Gaussian distribution. More marker screenings will take place with these two algorithms after the collection of high resolution mass spectra with the Waters Xevo instruments.
Tumor phenotyping
The objective of the task was to establish mathematical correlation between the mass spectrometric data and the status pattern of these well characterised markers. Furthermore, since detailed histological data was also available for all the tumours, correlation analysis was also performed in those cases. While the correlation between mass spectrometric data and cancer classification is highly important, the most interesting and innovative result of the current research will be the correlation analysis with the long-term patient history
The tumour phenotyping task was performed using breast cancer and colorectal cancer and by the 19th month of the project the breast cancer results are available. Since breast cancer is one of the best characterized human malignancies, it served as an ideal model. The tumour phenotyping task focussed on establishing a mathematical relationship between the composition of the REIMS(or DESI)-accessible complex lipidome and the clinical behaviour of the disease. Due to the limited timeframe of the project, the lipidomic information was correlated with known histological markers having validated diagnostic and prognostic value, for instance receptor expression status or histological tumour type. The results of this correlation analysis clearly proved that practically all tested histological parameters (and thus the phenotype of the disease) can be determined using the complex lipidomic fingerprint of the tumour with sensitivities and specificities well exceeding 90 and 97 %, respectively.
summary for WP5
Several algorithms have been tested to identify marker fingerprints specific on tumor tissues. The marker screening activity will be continued as more data entries with appropriate histopatological classification will be generated along with the data collection. The tumour phenotyping task was performed using breast cancer and colorectal cancer and by the 19th month of the project the breast cancer results are available. Since breast cancer is one of the best characterized human malignancies, it served as an ideal model. The tumour phenotyping task focussed on establishing a mathematical relationship between the composition of the REIMS(or DESI)-accessible complex lipidome and the clinical behaviour of the disease. Due to the limited timeframe of the project, the lipidomic information was correlated with known histological markers having validated diagnostic and prognostic value, for instance receptor expression status or histological tumour type. The results of this correlation analysis clearly proved that practically all tested histological parameters (and thus the phenotype of the disease) can be determined using the complex lipidomic fingerprint of the tumour with sensitivities and specificities well exceeding 90 and 97 %, respectively.
(WP7) Impact and IP protection
Objective
As the aim of the project was to further the development to a stage where the technology already makes commercial sense. The goal was to inform and if possible to persuade potential future users and decision-makers about the advances of this novel technology.
Existing IP needed to be harmonised and prosecuted and if possible novel protectable IP created.
Outside experts needed to be informed about the technology and included in the project.
Dissemination and Workshops
The articles published, demonstrations and workshops held are summarised in the appropriate forms.
We trust the targeted audience has been reached. Scientific publications were directed at the scientific community in high impact factor papers, and we managed to reach the general public through appearances in the BBC news in the UK, Spiegel in Germany, a large number of local media in Hungary. Outside the participating countries, the project, amongst others, was covered by Chinese (China Daily), Middle eastern (AlJazeera) and US (WSJ, Bloomberg, Cnet). As a result a considerable interest was generated both among potential investors to the technology and potential cooperating partners from the medical and mass spectrometric community as well.
Follow up demonstrations were performed to both media and academic institutions widening the scope of the project both in terms of technology and geographical penetration.
Patent applications
In the project plan we set out to harmonise our patent applications and to protect any new developments that may be protected. During the entire project we were working together with Knobe a US based IP law partnership to streamline and reinforce our IP portfolio.
The Jet Desorption Ionisation (JEDI) application has been transferred to Knobbe from Danubia, the local patent office. It is only the Hungarian patent prosecution of this application that has remained in Hungary for cost effectiveness. The transfer was required for quality reasons as the local office was unable to handle the fine details of the application language to the required standard. The cost of this transfer was mainly financed outside of this grant.
A JEDI spin-off created in the US in the form of a divisional application has been further pursued. The concept is to create an umbrella protection above the other – more technology oriented – applications, at least in the US for the concept.
The REIMS patent application was granted in Mexico, and has – after a round of clarifications and amendments – positive feedback from the US patent office.
The REIMS spin-off for liquid samples has been pursued further as it provides extra protection for the technology enabling the REIMS based analysis of biological fluids. The application has entered national phase in a number of countries on 28.06.2014.
The last application we were keeping in our portfolio is the most technical and is trying to provide protection for the collision surface discussed in WP1. This is the application where we have been developing the most hardware related know-how during the project, but have not decided to try to turn it to protectable IP. This application has also entered national phase in a number of countries on 28.06.2014.
During the project the IP creation was mostly on refinements in already existing areas – such as the collision surface that was mentioned above – or in the development of algorithms and software code, that in itself does not constitute patentable subject matter. The created know-how is therefore mostly protected through the careful publication of the results and through internal project documentation.
Potential Impact:
The overall and ultimate purpose of our project was to enable the ISD device to become a diagnostic tool in the hands of practicing physicians, making their work easier, quicker and more effective. By reaching the market and becoming commercially applicable in the various fields of medicine, oncology, surgery, microbiology and gastroenterology, ISD technology could introduce a completely new paradigm into histopathology. That of objective information used at multiple locations simultaneously. This can result in harmonized diagnosis, prognosis and treatment methods leading to a more globalized medical community opening novel pathways.
The concept of the Intelligent Surgical Device was conceived to provide oncosurgery a tool that – without the need of modifying current surgical techniques – is capable of providing information to the surgeon that was previously not available in the surgical environment. By combining profiling mass spectrometry with the standard surgical toolset, we can provide information to the surgeon about tumor margin and recidive location. As a result the surgeon is able to make a more valued decision quicker, decreasing time spent in the operating room and potentially decreasing the need for repeated operations.
The target of the current Grant was to develop this technology to a status where it is accepted by the scientific community. The technology was developed to a status where under R&D conditions (the aim was specifically not to try to create a commercial clinical study ready prototype, but to prove) we could collect sufficient data to provide basis for the above claims. A Science: Translational Medicine article in June 2013 (“Intraoperative Tissue Identification Using Rapid Evaporative Ionization Mass Spectrometry” Julia Balog, …, Zoltan Takats) about the technology created huge public interest with demonstrations on the newly developing machinery to major newspapers and television channels worldwide. The publicity created resulted in a number of leads for potential cooperations from the scientific community as well as the medical community and potential industrial partners.
We started talks with a large number of medical centers worldwide for potential co-operations in data and experience collection with our devices, but due to our limited resources (multiple dedicated specifically modified mass spectrometers are required at each potential location) we are still concentrating on maintaining our capabilities within the consortium hospitals and in cooperating parties in Hungary.
We also received considerable interest from medical and analytical device manufacturers to collaborate for project execution. The most welcome of these was that of Micromass UK, a mass spectrometry company. Micromass/Waters devices were the 3rd type of devices we have been lately integrating to the project after using mass spectrometers from Thermo Scientific (US) and Tofwerk (Switzerland). Seeing the future development requirements in the project Micromass UK proposed that it is willing to finance the project further though the development and later in the clinical phase in return for ownership in the IP.
After a long negotiation phase it was acknowledged, that it is not possible to both continue the FP7 funding and use the resources of Micromass UK and have – in order to ensure the long term continuation of the project – opted to ask for an early termination of the current Grant. This, we believe ensures that the project receives the backing it requires in the long term future without further need for funding form the Taxpayers while keeping the project within the European Union. R&D activities are planned to be distributed between Hungary and the United Kingdom with a considerable increase in personnel. The number of data collection size is planned to be increased considerably in the next couple of years with more participating doctors and drawn in from an increased number of fields of oncology. The microbiology and food safety scope of the technology is to be increased dramatically in the next few years as well.
As a consequence we strongly believe that this SME startup led FP7 financed project has fulfilled its main targets – although some of the technical objectives have, admittedly, not yet been met – in that it developed the technology and the database to a level where its achievements supported its commercial viability to a level where industry funding was achievable.
List of Websites:
www.medimass.com
The objective of the collaboration research project was the development of a mass spectrometry based tissue identification system capable of in-vivo, real time identification of tissues during surgical and diagnostic interventions. The underlying Rapid Evaporation Ionization Mass Spectrometry (REIMS) method has been developed prior to the project launch and according to the early feasibility studies it provides enough information for tissue characterization. In order to prove true intrasurgical applicability for a number of areas, a device interface, the surgical tools had to be adapted and extensively tested in real life situations.
In accordance with the objectives development of an advanced REIMS atmospheric interface for Xevo G2-S mass spectrometers (Waters) has been completed. Modifications on the benchtop instrument have been carried out in order to make that operating theatre compatible (surgical version). Three unit of this surgical versions for in vivo demonstrations and two additional benchtop G2-S for data collection have been delivered and installed. Electrosurgical handpieces with fume aspiration capabilities has been developed to support the demonstration and data collection activities. Addressing the data management needs separate software platforms for data collection and algorithm development has been developed.
Protocols for ex-vivo data collection and in-vivo applications have been completed, ethical permissions for the study in two locations have been granted. Data collection has been launched and in progress.
Results of preliminary studies in different areas of developments as well as representative data of applications in different surgical areas have been published extensively.
The main aim of the project was to develop the technology to a status where commercial applicability is no longer a distant possibility. The target was to develop the technology to status where data collected for validation is as close to the future application setting as possible for an R&D activity. The Consortium members achieved the above interim objective by completing most if not all the milestones described in the DoW, and were continuously working to ensure the long term survival of the project by proving the project’s ultimate commercial viability.
As a consequence of the first major communication round in June 2013, a number of 3rd parties indicated their willingness to cooperate with the project. Interest was recorded from both academic institutions and industrial partners. The decision was met to pursue these leads to try to create a long term strategy for the project. Towards the end of 2013 Waters UK expressed its interest in a closer cooperation which the consortium dully reported to the EU. In the discussions about preferred structure during the half of 2014 it became evident that the EU funding needs to be terminated if the long term funding of the project is to be secured through an industrial partnership. The EU was informed again and potential solutions discussed. We trust that this decision will secure the long term future of this most successful project, in which the Consortium Members agreed to carrion as defined in the original project plan irrespective of funding sources.
Project Context and Objectives:
The general aim of the proposed project is to show the commercial viability of the recently developed REIMS tissue identification method raising the technical status to a level where it could be introduced it into clinical environment obtaining sufficient amount of data to prove that a later regulatory process is worthwhile. The target of the R&D activity was to raise the technical status of the project to a level where it can grab enough global awareness to be able to secure the survival of the technology.
REIMS-based tissue identification is implemented by the on-line analysis of the aerosol formed on the electrosurgical dissection or coagulation process. MediMass Ltd. has been developing REIMS technology and corresponding instrumentation and bioinformatics toolkit for 3 years. Pre-clinical and clinical experiments revealed that the concept of real-time, in-situ, in-vivo tissue identification by REIMS technique is feasible, however standardisation and development of the device and large-scale clinical studies are needed for the accurate definition of the role of this phenotyping approach in the hospital patient journey.
The developed mass spectrometric devices will provide mass spectrometric chemical profiles of corresponding tissues, information which in itself cannot be interpreted by medical professionals. Furthermore, since the profile data does not carry “absolute” information, it can only be interpreted by comparison with a set of authentic data. Comparison will be implemented by a combination of unsupervised and supervised statistical multivariate pattern recognition algorithms. That is, instead of comparing unknown spectrum to large set of known entries, the database is converted into a histologically assigned data space, where the individual spectra represent single data points. Unknowns are identified by localizing the new datapoint on this data map, measuring its distance from surrounding data clouds (each representing a type of tissue) and classify it into the closest. This strategy allows practically real-time analysis, since the database conversion is performed prior to medical intervention. As it was already pointed out, the identification strategy calls for a large set of authentic data. One of the most important objectives of the proposed research is the construction of this database, however not only with classical histological assignment.
We planned to collect all relevant biochemical (proteomic and genetic marker statuses) and clinical (survival, drug sensitivity, metastatic behavior, aggressiveness, etc.) information and store it with the spectral information in an SQL-type database structure. This database allows the deciphering of correlations between our lipidomic/metabolomic data and status of traditional markers, and definition of novel, combinatorial biomarkers for the identification of “classic” phenotypes. Availability of clinical data and potential follow-up of patients allows the establishment of correlations between the characteristics of disease and tissue lipidome/metabolome. Well-established correlations (e.g. between abundance of certain complex lipid and tendency for metastasis formation) enable us to define novel predictive markers, and eventually a new type of (i.e lipidomic/metabolomic) clinical phenotyping system. We believe that small molecular markers give more comprehensive information on the short and mid-term status of the body than genetic or proteomic markers. Furthermore, these species also provide information on extragenomic factors such as- the composition and effect of associated microbial flora, which is a critically important factor in case of the diseases of gastrointestinal or urogenital systems.
The first step of proposed project was to establish the technical conditions for testing the methodology in clinical environment. Instrument development activities comprised the development of electrosurgical handpieces with auxiliary consumables, ion transfer devices, operating room compatible instrument chassis and so-called mass spectrometric front-end solutions. Development of necessary instrumentation is described in details in chapter “Development of dedicated mass spectrometric instrumentation for use in human medical environment”, Work package 1.
After providing the appropriate instrumental background the next step was the introduction of a mass spectrometric tissue identification device into operating theatre and optimizing the method to facilitate the translation of this new technique into standard operation practice. The clinical component of the proposal was also divided into two distinct work packages. One (Work package 2) focused on the practical utilization of the technique in surgical environment and collection of limited amount of in-vivo data (limited by ethical constraints), while the other,(Work package 3) comprised the large-scale acquisition of histologically authentic data using fresh ex-vivo samples. The in-vivo activities, grouped in Work package 2, targeted the collection of valuable feedback of surgeons from the different surgical areas regarding the special features related to the applicability of the REIMS based tissue identification method. The aim of the ex-vivo data collection tasks in Work package 3 was primarily a systematic collection of REIMS data from a range of tissue types in order to populate spectral databases for the effective training the recognition method. The larger spectral library used as a training set for a surgical area the better in-vivo classification efficacy can be achieved. While the full spectral library collection from the ex-vivo samples (3rd deliverable of WP3) provides a different aspect for understanding the fine structure of the tumours and its close proximity, utilizing the fine spatial resolution of the imaging techniques.
Work package 4 comprised development tasks targeting more universal biological tissue identification. The first task aims at a development of a spatial logging system by linking mass spectrometric data with spatial information using two different approach. The first (Simultaneous Camera Report, D4.1) targeted the exploration of the possibilities to assess spatial information by application of cameras around the area of intervention and storing the MS identification results together with the video captures in a time-synchronized manner. The other ( Spatial Logging System, D4.2) relied on instrumental background already available in some surgical environment where navigation devices are extensively used (such as neuro-navigation system in case of brain surgery). The simultaneous use of navigation with REIMS tissue identification, especially together with pre-operative imaging results, on one hand helps the surgeon to assess the spatial shift relative the pre-operative conditions and also give an indication where Mass spectrometric identification is needed (i.e. close to tumor border). The second task of this Work package (Spectral Interconversion Algorithm, D4.3) has the objective to provide an tool to convert spectral information collected with different mass spectrometers and/or different ionization methods. Such a tool could facilitate the deeper penetration of the scientific community as the currently instrument and ionization technique dependent datasets would become comparable.
The goal of the tasks collected in Work package 5 was to create a robust and stable bioinformatical toolkit for biological tissue identification and pattern analysis. Besides the improvement of the currently applied PCA/LDA based tissue classification algorithms, targeted engines for universal marker screening on the collected dataset have been planned to develop. The ultimate goal of this study is to identify new metabolomics/lipidomic fingerprints for tumor phenotyping and ultimately improve the clinical outcome in patients with malignant disease.
WP7 accumulated tasks that were targeted towards the long term viability of the R&D activity that is dissemination activities and IP maintenance. The goal was to raise sufficient awareness – both scientifically among the general public –, and to consolidate the IP portfolio such that it would form a basis for a long term product development.
Project Results:
Development of dedicated mass spectrometric instrumentation for use in human medical environment (WP1)
Objectives and introduction
The general objective of this work package was the development of instruments for Rapid Evaporative Ionization Mass Spectrometry (REIMS) based intraoperative tissue identification technology, which is suitable for use in the medical environment. We planned to develop electrosurgical intervention devices (handpieces, endoscopic snares…) that are capable to collect and transfer the surgical fume to the inlet of the mass spectrometer as well.
To provide an overview of the context of the tasks within the work package, after a short introduction of the ISD device and its accessories, the main functional units and their connection to the deliverables will be presented beforehand. Afterwards, a high level summary will be provided on the task level progress.
The device converts the chemical information of the surgical fume arriving from the area of intervention through the handpiece into mass spectra, which can be used for the characterization and identification of the tissue. The surgical fume is transferred to the atmospheric interface by a venturi airjet pump embedded in the venturi chamber attached onto the front plate of the atmospheric interface. The surgical fume is then driven to the atmospheric interface through a sampling capillary by the aspiration of the vacuum pumps of the mass spectrometer. The droplets and particles containing pre-formed ionic species then will be declustered assisted by collisions on the surface (of optimized geometry and position) and evaporation (due to low pressure and elevated temperature) resulting in ions that can be subject to separation according to their mass/charge ratio in the mass spectrometer. The acquired raw data are transferred to the database managed by Medimass/ISD SW. There a new model can be built using the histopathologically validated data sets, or in case of surgical applications, the resulting spectra can be used for real-time tissue characterization based on a model prepared earlier.
The device contains the following functional elements/units (1. Figure):
Handpiece
This is an electrosurgical intervention device, which has the functions of general electrosurgery, and is in addition to the standard functions suitable for sampling aerosol generated by the electrosurgical device for REIMS analysis. (see more in the report for deliverable 1.3)
Venturi chamber
This is the embodiment attached to the atmospheric interface equipped with a venturi airjet pump to transfer the surgical fume generated from the area of intervention through the handpiece into the atmospheric interface. (see further details in deliverable 1.1)
Atmospheric interface
This unit serves as an interface between the atmospheric pressure region of the venturi chamber and the vacuum region (of approx. 2 mbar) of the first zone of the ion optics of the mass spectrometer. Since this is the area where the preformed ionic species will be declustered from the droplets originally formed in the area of the surgical intervention, the design has a significant role in the generated ion intensity and as such, the overall sensitivity of the instrument. (see further details in deliverable 1.1 1.5)
Mass spectrometer
This is the unit that analyzes the generated ionic species according to their mass/charge ratio. For the surgical version, the original device will be redesigned to be compatible with the special requirements that originate in the application. (see further details in deliverable 1.1 1.5)
Software and database
Beside instrument control, raw data acquisition and pre-processing, the software had to answer all requirements that originate in the task of tissue characterization. Pattern-based tissue identification requires a database in which the tissue-specific spectra are stored (that are collected during the validation process), and a classification algorithm (real time recognition). The development of a software that covers all these needs and is able to handle different mass spectrometric platforms was indispensable as well. (See further details in the reports for deliverables 1.1,1.2 an all WP 2.5)
ISD prototype using existing MS platform
Operating room compatible (surgical) version development
The challenge was to convert a Waters G2S benchtop mass spectrometer such that it could be used in an operating theatre. To facilitate the transfer between the maintenance area and the theatre, the instrument had to be mobilized by proper application of handles and wheels (Figure 2.).
The cooling air circulation had to be redesigned in a way that the exhaust air could leave the chassis through a HEPA filter, so preventing the air of the operating room from being contaminated. This was accomplished by the attachment of a filter box to the chassis containing an exhaust fan and a series of filters and appropriate pressure sensors to monitor their performance as shown in 3. Figure and 4.
Atmospheric interface development
Cold collision surface
As a first step, we explored the effect of the collision surface geometry and position on the signal intensity. After a series of experiments with different shapes sizes of the collision surface, we concluded that the spherical collision surface with a given diameter positioned at a specific distance from the sampling capillary provided the optimal ion intensity. In order to do away with potential memory effectsdue to depositing on the collision surfaceheating was introduced to the collision surface. (5. Figure)
Heated collision surface
It was soon recognized that for the elimination of the depositing, a relatively high surface temperature is needed. The maintenance of this high temperature at normal operating conditions (continuous gas stream from the sampling capillary) requires quite a high heating power concentrated on in a relatively small volume. As the conclusion of a series of experiments, we settled with the application of a coil made of Kanthal strip, operating as a resistive heater at a high surface temperature at normal REIMS operating conditions (6. Figure).
Apart from preventing deposit from building up, the high collision surface temperature proved to be advantageous for the declustering process as well. A magnitude of intensity gain could be achieved in the REIMS spectra by heating the collision surface (7. Figure).
In conclusion an atmospheric interface for the Waters Xevo G2S mass spectrometers that provides intensive and stable signals for tissue recognition has been successfully developed.
The instrument has been converted into a device that is compatible with a surgical environment. Three of the surgical version instruments have already been installed and are being currently tested in a surgical environment. Apart from these devices, two original bench top Xevo G2S instruments will be used for data collection.
Software development
The task related to the software background can be divided into three main groups.
Instrument control
As the aim was to develop an instrument that does not require special expertise for use and maintenance, we decided that the instrument control functions have to be handled by the ISD software. Thus most of the tune parameters (to set the elements of the ion optics) will not be available for the end users but will be set by the system administrator for the optimal value. With this apart from the ease of use, we can assure that the acquired data will acquired with the same tune settings. To achieve this aim an interface program layer has been developed in close cooperation with the manufacturer.
workflow related development
Based on the feedback on the previous studies several suggestions and ideas emerged to improve the safety of the data collection and the accuracy of the histopathological classification by more strict sampling protocol. Implementation of the changes in the data collection protocol into the software has been decided. A user interface that follows the protocol step by step has been planned and its development has been launched. Additional features like handling and storing images of the samples before and after REIMS sampling and a user interface for assignment of the actual REIMS sampling point and the histopathology of the surrounding tissues will be implemented as well.
classification algorithm and data processing
As the currently applied mass spectrometric platform provide higher resolution data having quite different characteristic that the data structure we used before, significant changes had to be, and still have to be implemented to our data handling toolkit. Besides the benefits of the high resolution data collection (like separation of analytes with the same nominal molecular weight but different elemental composition, which provide a deeper view into the fine structure of the lipidomic fingerprint) it is more sensible to external parameters, like temperature and its effect on the performance of the elements of the electronics. To handle this sensitivity, a continuous adjustment of the instrument calibration using a selected molecular peak (Lock-mass) is required to maintain the resolution through the measurements. Apart from this feature we identified tasks around the noise filtering as, according to our experiences so far, we are confident that some of the periodic noise we experienced only with this mass spectrometric platform can be separated from the signals containing the chemical information, thus making the tissue specific differences more visible. To support the development of such a noise filtering algorithms development of a separate software platform (Offline Data Mining (ODM)) has been launched. Further results and details on these developments can be found in the chapter about the last task in this work package (Finetuning the ISD prototype, further step towards a medical device) or in the deliverable report (D1.5)
REIMS imager device for ex-vivo sampling
The objective of the whole project is the development of real-time tissue identification method that can be applied during surgical and diagnostic intervention. The tissue identification is based on the comparison of unknown data collected real-time to a database of authentic REIMS data by utilizing multivariate statistical pattern recognition algorithms. Therefore the applicability and performance of the tissue characterization is highly dependent on the amount of the histologically validated data points in the different application areas stored in the database. In this section the summary of the development of an automatized device, capable to high-throughput data collection, will be presented.
REIMS compatible imager platforms have been developed in two phases via the conversion of commercially available imaging stages. The first version, which meant the modification of the Prosolia DESI stage specifically purchased for this purpose by MediMass, was capable of collecting data from samples from even surfaces like tissue sections. This has been tested and compared to DESI imaging. Based on the initial experiences, an advanced imaging stage has been developed via the modification of Prosolia Flowprobe configuration (8. Figure), which provides better conditions for data collection from larger tissue specimens (which usually have uneven surfaces) since the probe positioning along the z axis is continuously adjusted according to a laser distance sensing (purchased separately at Imperial College).
For the REIMS data collection the original Flowprobe sampling head has been replaced with an electrosurgical blade holder (9. Figure). This REIMS sampling head is equipped with an aspiration channel for aerosol sampling and a replaceable blade holder (connector for tubular probe needle) and is designed to facilitate easy maintenance and decontamination. The standard tubular probe needle is similar to the needle electrode that is used during standard electrosurgical interventions.
Due to the various safety features incorporated to protect the patient, the commercial electrosurgical supply does not easily lend itself to an automated analysis platform, thus the use of a conventional power supply simplifies the experimental setup. The variations of the spectral outcome due to the application of different power supplies have been tested and found negligible.
As a result of this task a robust REIMS imager with excellent precision has been developed. For data collection a mixed application of manual and automated methodology has been considered so far. While the manual data collection has a limited throughput and also provide some more possibilities for human dependent variability, the high-throughput automated approach may contain technical features that may result in quite another outcome than the real life application. Therefore, for a proper benefit analysis a comparative study has to be executed on a larger data set.
REIMS compatible interventional devices
As electrosurgical methods are widely used during surgical interventions in different areas, the variety of the available interventional devices (handpieces) reflects the diverse needs of the different applications. Supporting the introduction of the REIMS base tissue identification method into clinical practice, universally applied handpieces were selected for the development.
Monopolar handpiece
Since the monopolar electrosurgery is one of the most generally used techniques in surgery, the monopolar handpiece with cut and coagulation functions was first selected to modify for surgical fume sampling. Based on the surgeons feedback we decided to supply four exchangeable electrodes for the different applications a needle and a blade of two different lengths of both. For the development and production of the sampling handpieces, the collaboration with Medres Ltd. has been continued. They have considerable experience in the field of medical device development. At present, development PHASE I has been closed, project documents, manufacturing drawings and instructions, material specifications and preliminary instruction for use have been finalized.
As a result of the progress handpieces for the ex-vivo manual data collection and demonstrations can be provided in sufficient quantity, and sterile handpieces in appropriate packaging will be available soon for the in-vivo demonstrations and validation tests (10. Figure and 11. Figure).
Bipolar forceps
Bipolar electrosurgery is a commonly used method of surgery as well. The device introduced to the intervention area consists of two electrodes separated by insulation. The electrical current only passes through the tissue between the two electrodes. The advantage of the bipolar technology is that the distance between the electrodes of the device (forceps stems) is only a few millimetres, therefore only very small sections of tissue are involved in the electric circuit. As opposed to the monopolar devices, bipolar forcepses are not applied for cutting, but much more for coagulations, since these use lower current density due to the much smaller tissue resistance. Bipolar forcepses are frequently used in neurosurgery, as they are precise, less invasive and stay current-free.
After a thorough search among the available bipolar handpieces, a type of forceps was found that seemed to be suitable for our application. This irrigation forceps has a built-in tube and an opening near its end (12. Figure), which is normally used for dispensing irrigation liquids to the surgical site. We have found that this type of forceps is applicable for aerosol sampling using the irrigation channel for the surgical fume sampling. As the amount of the generated aerosol is much smaller, due to the less invasive nature of this technique, the fume transfer tube is connected directly to the sampling capillary of the mass spectrometer and the aerosol is aspirated by the vacuum pumps of the mass spectrometer.
Sampling probe
During the consultations with the surgeons the need emerged for a sampling probe that is especially designed for data collection only without features supporting intervention processes like cutting. These devices could be effectively used to analyse the residual tissue layers on the wall of the cavity after the removal of tumours to ensure that the margin is clear. As this probing concept could be effectively utilized as a diagnostic approach, feasibility studies have been launched to explore its potential benefits.
To summarize the results in this area, we can conclude that the development steps required for the mass production of a disposable monopolar handpiece have been dully completed. Manufacturing phase IIa has been closed, so a critical milestone towards the transferable design has been achieved. The required amount for the data collection activity is available. The sterile end products will be available soon for the in vivo demonstrations. The bipolar applications can be supported with commercially available intervention devices. Preparations for probe development have been launched.
Adaptation to endoscopy
The main objective of the endoscopy sub-project was to couple the REIMS technology with endoscopy and produce a device capable of the in-vivo testing of gastrointestinal lesions in course of diagnostic interventions.
Endoscopic snare
The snare device comprises a hand tool, which is used for the opening and closing of the snare (by pushing the loop out of the endoscopic device and pulling it back on) and a long stainless steel thread which ends in the snare loop itself. In order to isolate the snare from the body of the endoscopic device, the stainless steel thread and the loop are surrounded by a plastic tube housing. Since the outer diameter of the plastic (HDPE or PTFE) housing is slightly smaller (<0.2 mm difference) than the internal diameter of the working channel of the endoscope, the electrosurgical aerosol containing charged particles needs to be aspirated through the housing, as an ion transfer device. There are number of endoscopic devices available with multiple work channels, however aspiration through an alternative channel was proven to be less efficient due to the inherent distance between channels and also due to the interference with the optical elements (camera, light source). Employing the snare housing as an ion transfer device has raised a number of problems including the blocking of the housing by the tissue being dissected, as it is shown on 13. Figure .
When a polyp is captured by the endoscopist, its basal part is pulled against the housing before the actual dissection takes place. As a result, the aerosols formed on dissection cannot be aspirated through the housing, since its entrance is completely blocked. Further problem with the snare housing was associated with its connection to the Venturi air jet pump-based aerosol aspiration system, since the housing tube has a dead end on the outer (handpiece) end of the device. The potential neutralization of charged particles on the bare metallic surface of the stainless steel thread was perceived as an additional problem. In order to overcome the second problem (i.e. connecting the snare housing to the aerosol aspiration system of the mass spectrometer) and also to minimize the extent of modifications made to commercially available, fully approved snares, a snare equipped with irrigation capability (manufacturer) was used as an initial experimental setup. Since the irrigation function requires the flushing of the snare with physiological saline, these snares feature a Luer port on the outer end of the device, which can easily be connected to the existing aerosol aspiration system. Due to the potential risk of short-circuiting the snare with the flushing liquid, the snare surface is covered by a polymeric varnish film in this device, which gives an efficient solution for the last problem mentioned above (neutralization of charged particles). The problem of blocking the housing inlet with the sample itself could not be solved without modifying the device. In this case the wall of the housing tubing was fenestrated close to its distal end. The shape, pattern and number of fenestrations was carefully optimized to make a compromise between optimal aerosol sampling efficiency and the mechanical stability of the plastic tube housing. The resulting slightly modified endoscopic snare was connected to the modified Xevo G2-S surgical mass spectrometer system and proof-of-concept data was collected using food-grade post-mortem porcine colon and gastric mucosa as samples.
Special accessories for endoscopic applications
Since the endoscopic application of the electrosurgical intervention occurs in cavities of the human body, the applied fume aspiration from these areas has some special features. To prepare for those the following additional hardware elements have been developed.
Triggerbox
As the intervention area has a finite volume, and the air resupply through the work channel of the probe is also limited, the continuous aspiration of the gas should be avoided in most of the cases. To address this - although it was not mentioned among the tasks of the grant - we have developed a device which monitors the operation of the electrosurgical handpiece, and depending on this, is capable to provide trigger signals towards external devices like the bypass valve attached to the exhaust of the venturi chamber, the gas supply of the venturi air jet pump and the mass spectrometer. With this synchronization, only for the duration of the actual intervention, both the fume aspiration and data acquisition be switched on. To provide flexibility for the users, the triggerbox can operate in continuous mode as well, then all the triggers are switched on independently on the status of the handpiece.
Liquid separator
Surgical handpieces equipped with aerosol aspiration systems tends to aspirate not only aerosol but also liquid accumulated in the surgical area. Aspirated liquid interferes with ion transfer and can seriously damage the mass spectrometer, so to handle this problem a route scouting of the development possibilities of a liquid separation device has been carried out. Several approaches have been considered, prototypes (1. optical droplet sensor drived valve system, 2. continuous centrifugal separator, 3. absorption based separator and a 4. conventional liquid trap) have been manufactured and tested. Two of the solutions (1 and 2) can be used in continuous mode, which is a significant benefit as they do not require additional care to change the buffer. At this stage of development we can say, that the most innovative technical solutions (1-3) are either too complex to provide a proper base for a development of a disposable or sterilizable accessory or still suffer from compromised operational reliability (1). Therefore we have to iterate further to find more simple but effective designs capable to operate in continuous mode. On the meantime we use the conventional liquid trap and special attention will be paid to emptying the buffer in time.
In conclusion, the REIMS technology was successfully coupled with both gastroscopy and colonoscopy devices. The resulting iEndoscope setup features mostly commercially available, fully approved medical devices, at least regarding the parts used in the close proximity of the patient. This fact guarantees that the instrument can enter into formal data collection without any additional regulatory approval procedure. The special requirements of the endoscopic environment has been properly addressed by development of liquid separator and triggerbox device.
Finetuning the ISD prototype, further step towards a medical device
This task was defined to collect all activities around fine-tuning the system pre-prototype obtained according to deliverable 1.1. The developments were primarily driven by the user feedback from surgeons, pathologists and research fellows, but some of the new features are just the natural implementation of the new ideas targeting the organized improvement of the system and its SW background.
Hardware modifications, atmospheric interface
Isolator plate
During the risk assessment procedure, minor modification requests emerged with respect to the atmospheric interface of an operating theatre compatible mass spectrometer under the following hypothetic fault condition. The surgical fume transfer tube gets filled with conductive material (saline) and thus creates galvanic contact between the patient and the interface chamber. As the interface chamber is connected to the technical earth, while the patient is connected to the patient earth, due to the possible potential differences between the two earth circuits, current might be generated with all its disadvantageous consequences.
To mitigate this risk, the interface chamber has been separated from the mass spectrometer with an isolating plate, and a patient - earth connection has been installed onto the interface chamber Therefore, under the fault condition described above, the separate earth circuits remain unconnected.
Integrated power supply and temperature control of the heated coil
For the simple and convenient setting of the atmospheric interface parameters (coil offset voltage and coil heating current), the external controlled current power supplies have been replaced with an alternative solution involving the available supply channels on the mass spectrometer. Besides producing simpler thus safer equipment, the coil offset setting from this point could be adjusted from the mass spectrometric tune page, and also could be logged to the experimental file in the acquisition folder.
In order to find a truly robust solution is to close the control loop with a sensor capable of capturing temperature on the kanthal coil surface and modulating the supply current according to the measured values, investigations have been carried out to find a proper method for temperature sensing. As a result of this development cycle, a prototype of a closed loop coil heating control circuit has been built based on the temperature sensing by a remote IR sensor (14. Figure).
So at this point, we can state that we have a backup solution at hand for the coil temperature control, although the integration to the mass spectrometer is still ahead. In the event of the controlled voltage power supply functioning adequately witnessed by the different users at various locations, that version will be used. In order to provide a backup solution, however, we will also continue our experiments on the closed loop temperature control approach
Software development
During the early data collection campaigns, a quite a number of important experiences was collected, which initialized the development of a new data collection module for the ISD Software package. The aim of this development is to improve the quality of the data collection by decreasing the possibilities of the human- and workflow-related errors by the application of barcode labeling and handling, and to provide further interfaces for more accurate referencing between the tissue images and the mass spectrometric data acquisition. New features have been introduced as well to support the more detailed histopathological validation and its better organized administration.
Along with the increasing amount of collected data from the different areas, a need has arisen for a support software platform, where new algorithms for the different phases of the data processing (noise filtering, background removal, feature selection, etc) could be conveniently tested and evaluated without disturbing interference with the data collection. For the demonstration of the role and performance of this offline data mining platform (ODM), please see the results of an algorithm for the selective removal of a periodic chemical noise from the mass spectra on 15. Figure.
Handpieces
To summarize the results in this area we can conclude that the mass production of a disposable monopolar handpiece has been organized. Manufacturing development phase IIa has been closed, biocompatibility tests have been specified. For the sterile product development the dose map protocol has been finalized. The validation tests were always planned to be outside the scope of the EU funded project, and will be completed during the latter part of 2014.
The required amount for the data collection activity is available. The sterile end products will be available soon for the in vivo demonstrations. The bipolar applications can be supported with commercially available intervention devices. Preparations for probe development have been launched.
As a summary, it can be stated the ISD system with all of its components become nicer safer, more integrated. The software background stepped forward towards a higher quality and more marketable product, and with the support of the ODM platform we have an effective toolkit to utilize further the tremendous amount of valuable array of information that was hidden in the forest of the mass spectral fingerprints.
Summary for WP1
To summarize the progress of the tasks belonging to this work package (WP1) all the elements of the technical background has been developed and the system is ready to support the data collection activities grouped in the other work packages. Development related to mass spectrometers including atmospheric interface and the area of operating room compatibility resulted quite matured solutions through several engineering iterations. We managed to either develop or identify commercially available interventional devices to support data collection in the targeted surgical areas. Special requirements related to Endoscopic applications have been properly answered. REIMS imager device to support automated data collection has been developed. Software development targeting the integration of the waters mass spectrometric platform and data processing protocols specially required for the high resolution nature of the applied mass spectrometric platform has been launched
Clinical development and in-vivo data collection (WP2)
Objective
The aim of this work package is the introduction of a mass spectrometric tissue identification device into operating theatre and based on the early user feedbacks optimize the method to facilitate the translation of this new technique into standard operation practice.
In-vivo data collection was planned to start in 2014. The inclusion of hospital units was always an issue for the project management (see the discussion in section 3.2.2) and due to the potential early termination the answer to this question was delayed. As a result Asklepios Medical School (AMS) played only an advisory role in the project (the doctors and R&D personnel have not been remunerated for their services), no formal link with the healing unit has been established, most of the work therefore has been allocated to institutions belonging to Imperial College. This resulted in that some of the areas, where the experiments were allocated to AMS (like brain surgery), got a bit behind the schedule due to the delays caused by the uncertainty whether the grant funding will be terminated early.
Optimal settings and in vivo protocols for surgical interventions
This task comprises activities to generate the necessary documents to facilitate the safe and reliable experimental work in the targeted surgical area including documents providing clear user instructions for the system applied and protocols to ensure the quality of the sample collection and handling.
Operation instructions for the available instruments
Two detailed user manual has been compiled for the two different ISD prototype containing all the necessary information to set up the specified operational parameters. One for the instrument based on Tofwerk CTOF mass spectrometer, and another for the waters Xevo-G2S based instruments. These documents can be found among the deliverables for this task (D2_1TofWerk optimal settings handling protocol, D2_1Xevo G2-S optimal settings handling protocol)
Protocols for in vivo operations
As there are significant differences in the practices and environment in the different surgical areas, besides a general protocol several other one had to be composed taken into account the special needs of the appropriate area. Apart from the liver brain and colorectal protocols most recently another one has been added to the collection to cover the breast surgery as well. Ethical approvals to carry out the protocols has been received for both locations.
Intraoperative spectra
The systematic collection of human colorectal, liver and breast tissue has been started in the past year and a half, while everything is ready for the kick-off of the collection of brain tissue. This latter is the one which is relatively , but mildlybehind the schedule. Theatre compatible surgical devices have been created and shipped to Imperial College for in the theatre use. These instruments have been used during liver, colorectal and brain surgery, and will be used in the near future for breast surgery also. Devices for Asklepios Hospitals are ready and will be shipped and put in use as soon the termination of the grant funding is finalised and a new structure is put in place. All planned units have received the necessary ethical backing for starting the procedure.
Breast
Breast cancer is the most common cancer affecting females in the United Kingdom and around the EU. In 2011, 41,523 women in the UK were diagnosed with invasive breast cancer the vast majority of whom received surgery with curative intent. Positive tumour resection margin (tumour at the inked margin) following attempted breast conserving surgery (BCS) is one of the most important contributing risk factors for ipsilateral local recurrence and it is recommended that in cases where margins are positive or “close” that further surgery is carried out to confirm that margins are clear of disease. A recent UK National Audit of screen detected breast cancers indicates that ~24% of women undergoing BCS require additional surgery to treat positive resection margins.
Our aim was to investigate whether near real-time tissue identification using REIMS is feasible for identification of normal, benign and malignant breast tissues (using formal histological examination as “gold standard”), and whether accurate predictions regarding margin status can be made based on REIMS data acquired in-vivo.
Our data suggests that the metabolic spectra of aerosols obtained from ex-vivo healthy and malignant breast tissues are sufficiently distinct to warrant more detailed investigation toward an in-vivo tool for margin control. All ethical permissions and clearance with the hospitals for in-vivo work were acquired, a theatre compatible, packaged device has been moved to Charing Cross Hospital and arrangement was made with breast consultants. We are continuing the ex-vivo breast tissue collection and the in-vivo work will launch n Q3 2014 in Charing Cross Hospital. We are also submitting a proposal for CRUK CTAAC in order to continue the work and answer the stage II questions.
Colorectal
Colorectal cancers are still one of the leading morbidities in US and Europe, while gastric cancers have very poor outcomes, thus an early diagnosis is an essential factor for long term survival. Rapid Evaporative Ionization Mass Spectrometry (REIMS) is a technique developed for the in-vivo characterization of human tissues by mass spectrometric analysis of the aerosol released during electrosurgical dissection. The ionization technique can also be combined with an endoscopic snare, allowing an in-situ, in-vivo analysis within the gastrointestinal lumen. The aim of this study was to develop and characterize endoscopic and colonoscopic setups, and to test the method and classification performance in-vivo. An ex-vivo tissue specific, histologically assigned spectra database is necessary for in-vivo studies, thus the first aim was to create a spectral database from ex-vivo fresh samples.
Both the monopolar and iEndoscopic setup was tested in-vivo during surgical resection of colorectal cancer or endoscopic resection of polyp in the colon. Both setups were suitable for in-vivo operation, thus the study will continue with the in-vivo phase.
Our results have shown, that there is a clear separation between healthy, cancerous and adenomatous polyp tissue spectra, as there is a clear separation between the different layers of the gastrointestinal tract. The results also suggest, that ex-vivo colorectal database is suitable for the identification of tissue acquired with our novel iEndoscopic setup. In-vivo experiments have been started, and will continue in the future for colorectal tissue and endoscopic interventions in St Mary’s Hospital and Royal Marsden Hospital.
Liver
Liver cancer is still a leading problem around the EU. Most of the patients have metastasis from other primary cancers, mainly from the gastrointestinal tract and lungs. Besides metastasis, primary liver cancers are also hard to detect, as by the time symptoms arise, usually a greater portion of the liver is affected. It has become more and more wide spread not to remove a full lobe in most cases, but to keep healthy liver parenchyma. In these cases our iKnife could be a very useful tool to detect the margins of the cancer or to identify small, unknown spots occurring in these patients throughout the liver parenchyma.
The instrument has been used in-vivo during surgical resections of liver cancer a number of times. Liver parenchyma contains a mass of blood, which could affect our results. The blood in the system does make the acquired spectra more noisy, however it did not affect the correct identification rate. In-vivo iKnife measurements were carried out in Hammersmith Hospital. In all cases, healthy liver parenchyma was sampled in the normal procedure while removing the tumour, and cancerous tissue was only sampled ex-vivo immediately after removing the tissue from the patient. All spectra were classified using the previously built ex-vivo tissue database. Our results showed >90% correct classification rate, however as healthy tissue was not removed from the patient and could not be validated by histopathologist, precise specificity and sensitivity data could not be calculated.
Brain
All brain experiments should started in Asklepios Clinic, Hamburg. The global ethical permission, including brain surgery, has been accepted now at Imperial College London, an instrument has been transported to the 14th floor surgical block of Charing Cross Hospital, and the in-vivo measurements will start around the last week of July. Our previous results collected in the Surgical Department of University of Debrecen both in-vivo and ex-vivo suggest, that iKnife coupled with bipolar forceps is capable of differentiating healthy, malignant and benign tissue both ex-vivo and in-vivo. Tissue was collected from various type of primary and metastatic cancers, benign and malignant tissue in a total number of 64 patients. Healthy tissue was collected from 6 patients, all during the surgical procedure in-vivo. The correct classification rate after leave one patient out cross validation was between 80-100%, however the number of healthy (n=30) and non-healthy (n=370) spectra has an order of magnitude difference, which biases all the results. In the future, more healthy and cancerous tissue spectra has to be acquired during surgery and immediately after surgery, still in the theatre.
Gastrointestinal surgical device
The main objective of this sub-project was to utilize the technical background developed in the first work package (see report for D1.4) for in-vivo testing of gastrointestinal lesions in course of diagnostic interventions. According to the schedule of the data collection activities, the endoscopic and laparoscopic results and in vivo demonstrations were going to start as of month 18. As we have accomplished all the technical development tasks, this part of the project is progressing in a timely manner. The produced data presented here and in the deliverable report all derive from ex vivo procedures; we are now in the position to launch the in vivo demonstrations and the data collection campaign in this field as well.
The final system was tested ex-vivo for human samples. Gastric cancer and healthy gastric mucosa samples were collected at the Department of Histopathology, St. Mary’s Hospital, Paddington, from complete gastrectomy specimens by histopathology consultants. Healthy gastric mucosa and gastric adenocarcinoma – similarly to earlier observations – were found to show markedly different spectral profiles as it is demonstrated by the multivariate statistical plot on 16. Figure.
In conclusion, the REIMS endoscopy device is ready for human testing, both from instrumental and ethical point of view. The experiments will start in July at St. Mary’s Hospital (ICL).
Summary for WP2
Most of the surgical area we originally planned to introduce this new methodology has been successfully involved into the cooperation. Although in some cases, especially in the area of brain surgery (due to abovementioned reasons), the expected numbers of case has not been achieved by the proposed time, we are confident that having passed the warm-up period the progress will be a bit accelerated.
High throughput data collection for the development of spectral database (WP3)
Objective
The aim of this work package was to establish the infrastructure and methodology for large scale acquisition of histologically assigned REIMS data set. In order to ensure the quality of the collected data compilation of a protocol for histopathological and mass spectrometric characterization of ex vivo human tissue samples was also indispensable.
Protocols for sample collection
A protocol has been compiled with all the details that is required to eliminate risks that may result diversification of the resulted data due to inappropriate sample handling. The workflow regulates also the sampling geometry to facilitate the safe histological assignation of the REIMS sampling spots. During the early data collection campaigns, a quite a number of important experiences was collected, which initialized the development of a new data collection module for the ISD Software package. The aim of this development is to improve the quality of the data collection by decreasing the possibilities of the human- and workflow-related errors by the application of barcode labeling and handling, and to provide further interfaces for more accurate referencing between the tissue images and the mass spectrometric data acquisition. New features have been introduced as well to support the more detailed histopathological validation and its better organized administration.
REIMS spectral library
As our attempt to launch the collection at Asklepios Medical School were delayed, we started all the tasks related to data collection to institutions belonging to Imperial College. The systematic collection of human colorectal, liver and breast tissue has been going on in the past year and a half at Imperial College, while everything is ready for the kick-off of the collection of brain tissues. Due to the difficulties around the data collection sites mentioned above, we did not manage to achieve the number of data entries planned for the 18th months, but we are still confident that after the successful full start of the data collection activity we would have sped up and proceeded in a timely manner. The number of samples collected is increasing each month, as more consultants are becoming involved, and new research nurses and clinical fellows are being hired.
Full spectral library
The task was aimed at the histological mapping of REIMS-accessible biochemistry with the rationale of validating the already described biomarkers (single or combinatorial) and systematic discovery of novel ones. At the first stage of the sub-project DESi-MSI was selected as the optimal imaging technology for the purpose. Since the chemical application range of DESI and REIMS show excellent overlap (both techniques detect predominantly complex lipid species), DESI does not suffer from experiment-related spectral artefacts and the technique provides sufficient sensitivity, the task was accomplished using this approach. Close to a thousand individual tumour specimens were analysed and over 2 million histologically assigned spectra were acquired. This unique database has revealed significant biochemical differences between tumours and surrounding healthy tissues and also clearly indicated that the tumour-specific biochemical alterations cross histological borders resulting in a well-defined tumour environmental biochemical segment. The observations pointed the attention to certain molecular species and their combinations and also gave evidence for the feasibility of proximal tumour detection by REIMS.
Summary for WP3
All the essential prerequisite infrastructure and methodology has been established for the large scale data collection. Though the number of collected data entries defined in the second milestone for the 18th months has not yet been achieved, we are still confident that after the successful full start of the data collection activity we would have sped up and proceeded in a timely manner. Regarding the full spectral library collection and its utilization to discover the fine structure of the tumors and their proximity we managed to achieve significant process ahead of the schedule.
Bioinformatics- spatial logging system and platform free development (WP4)
objective
In this work package tasks targeting the establishment of a platform for universal biological tissue identification has been collected. Development of a spectral interconversion algorithm to create possibility of universal application of mass spectrometric datasets generated with different ionization methods (laser, bipolar and unipolar electrosurgical devices) has been scheduled. A spatial logging system to correlate spatial and spectral data has been planned to be developed as well.
Simultaneous camera report
The object of the data recording is to correlate the mass spectrum with the exact location in the tissue where it was generated from. In some cases – for example during most brain surgeries – there is a spatial logging system or neuro-navigation which helps the surgeon with precise localization in the brain. However, in most abdominal and thoracic cases, there is no such navigation system available, the surgeon can only relies on his/her own eye-sight. Consequently, in these cases we need a simple localization system to validate our in-vivo data.
Originally, the implementation of the cameras was designed to occur simultaneously with the REIMS ISD data collection. Considering the fact, however, that the application of the ISD evokes quite a great deal of stress for the majority of the volunteering surgeons and their teams, we decided not to intensify the already existing pressure by involving new devices, such as cameras. Furthermore, the ethical issues arising from the process of embedding our methods in the local environment can also be handled more easily, if the modifications affecting the usual procedures are implemented in a gradual way. As a result, launching our validation experiments is deferred to a later date.
Proof of concept experiments has been performed to outline the possibilities to achieve the original objectives in this area. Feedbacks collected during these early studies convinced us that further efforts to develop a more sophisticated system should be deferred to a later phase of the project when the environment for in vivo validation studies becomes well established (expected by end of 2014
Spatial logging system
This sub project is aimed at the simultaneous use of navigation with REIMS tissue identification This combination on one hand helps the surgeon to assess the spatial shift relative the pre-operative conditions and also give an indication where Mass spectrometric identification is needed (i.e. close to tumor border).
We originally planned to carry out experiments targeting the integration of neuronavigation devices with the ISD technology at Asklepios Medical School. As our attempt to launch this collaboration failed, we had to reorganize the data collection activities to institutions belonging to Imperial College.
Due to difficulties mentioned above, this sub project is still in the preparation phase, thus a bit behind the schedule. Once we manage to launch the collaboration with the Neurosurgical Department of Charing Cross Hospital (Imperial College), which is planned in second half of 2014, the progress in this area will be more visible.
Spectral interconversion algorithm
Data fusion strategy to convert raw mass spectral intensities of biological samples measured by distinct desorption ionization methods and/or instrumental setups to cross-platform normalized analyte profiles has been developed. This strategy starts with the database driven analyte peak annotation and summarization, and after that, it derives analyte-specific normalization factors to adjust intensities of a common subset of analytes to a comparable scale between platforms. It is assumed that distinct MS-based platforms capture a similar set of chemical species in a biological sample, though these species may exhibit platform-specific molecular ion pattern distributions (17. Figure). The method was validated on a dataset obtained by laser desorption ionization (LDI), Desorption Electrospray Ionization (DESI) and Rapid Evaporative Ionization mass spectrometric (REIMS) analysis of porcine tissues. We demonstrate the capacity of our method to reduce MS-platform specific variation resulting in 1) high inter-platform concordance coefficients of analyte intensities 2) clear principal component-based clustering of analyte profiles according to histological tissue types, irrespective of the used desorption ionization technique or mass spectrometer and 3) accurate “blind” classification of histological tissue types using cross-platform normalized analyte profiles.
MS based tissue profiling and imaging techniques are beginning to translate into the clinical domain. In order to provide accurate chemical tissue identification, large histologically assigned databases, likely to be acquired over many years on multiple instruments and platforms, are required. The proposed XMS data fusion method offers a promising approach for converting raw mass spectral intensities of biological samples measured by distinct desorption ionization methods and/or instrumental setups to cross-platform normalized analyte profiles, We have demonstrated the capacity of the XMS method in order to sufficiently reduce platform/instrument specific variation in MS data resulting in high inter-platform concordance coefficients of analyte intensities, accurate classification of histological tissue types, measured on multiple platforms, irrespective of the desorption ionization technique or mass spectrometer used. The proposed validation strategy can also be used for further refinement of cross-platform normalization methods. The accurate tissue identification obtained on XMS normalized data, allows imaging-based desorption ionization methods to be used for the population of large scale MS-based tissue databases which are suitable for near real time intra-operative tissue identification using REIMS technology.
Summary for WP4
Proof of concept experiments has been performed to outline the possibilities for harvesting spatial information along with the mass spectrometric data. Feedbacks collected during these early studies convinced us that further efforts to develop a more sophisticated system should be deferred to a later phase of the project when the environment for in vivo validation studies becomes well established (expected by end of 2014). A functional toolkit has been successfully developed for interconversion of spectra collected with different ionization methods.
Bioinformatics- pattern and phenotype analysis (WP5)
Objective
The goal of this project is to create a robust and stable bioinformatical toolkit for biological tissue identification and pattern analysis. Targeted engines for universal marker screening on the collected dataset have been planned to develop. The ultimate goal of this study is to identify new metabolomics/lipidomic fingerprints for tumor phenotyping and ultimately improve the clinical outcome in patients with malignant disease.
Algorithm for marker identification
The objective of this task was to define different algorithms for the search of specific tumor markers in the human database. The algorithm we used so far is based on the complete spectrum as a fingerprint, however, in case of mixed spectra (part healthy, part cancerous tissue), monitoring specific molecule peaks could help in the precise identification of the sample. We have to point this out, even if there is only a small amount of cancerous tissue in the spectra. We have tried a couple of different algorithms: First, we tried to look at the loading plots of the PCA, however the most significant peaks were the most intensive peaks, thus we moved on to more precise algorithms among which the following three were selected for the systematic screen.
Biomarker m/z feature recovery with Bayesian analysis
An empirical Bayesian analysis of variance (ANOVA) was employed in order to determine the relevance of a given molecular ion species with regards to observed tissue type discrimination. This analysis is specifically suited for high-throughput biological datasets, as it borrows the information across metabolites for improved variance estimate of the null hypotheses of non-differentially abundant chemical species.
Group deviation filtering
As a first step we search for peaks on the m/z axis in the spectra which can differentiate between the various groups. These peaks will be assigned biomarkers by a probability. The perfect biomarker would be a peak with probability 1.
Kolmogorov-Smirnov nonparametric statistical test
This algorithm is used when the distribution of the data is unknown. It can be used on any kind of data and answers the question whether two populations belong to the same distribution or not.
systematic marker screening
The object of this task was to define different algorithms for the search of specific tumor markers in the human database. We have used the previously described algorithms on different spectra acquired from healthy and cancerous human tissue. The mass of the current human database was acquired with Thermo LCQ Deca or LTQ Velos mass spectrometers. These instruments are of low resolution, thus the dataset generated does not contain individual phospholipid peaks but the sum of 2-3 different species in one peak. The high resolution Waters Xevo G2-S instruments – more suitable to provide decent data for marker screening - were placed in the theatres the past month or will be placed there early next year, thus a large amount of data collection will start from January, 2014 as part of WP2 and WP3. As soon as at least 100 healthy and 100 cancerous human spectra of the same area is acquired, we will run these algorithms again on that dataset. We have tried marker screening on a small amount of high resolution spectra acquired with the instrument Thermo Exactive, and as a large amount of data is required for the algorithms to work properly, we also tried them on low resolution LTQ data.
When comparing the results, the algorithm based on simple ANOVA and the Kolmogorov-Smirnov test pointed out more potential “biomarker” peaks, than the algorithm based on the Gaussian distribution. More marker screenings will take place with these two algorithms after the collection of high resolution mass spectra with the Waters Xevo instruments.
Tumor phenotyping
The objective of the task was to establish mathematical correlation between the mass spectrometric data and the status pattern of these well characterised markers. Furthermore, since detailed histological data was also available for all the tumours, correlation analysis was also performed in those cases. While the correlation between mass spectrometric data and cancer classification is highly important, the most interesting and innovative result of the current research will be the correlation analysis with the long-term patient history
The tumour phenotyping task was performed using breast cancer and colorectal cancer and by the 19th month of the project the breast cancer results are available. Since breast cancer is one of the best characterized human malignancies, it served as an ideal model. The tumour phenotyping task focussed on establishing a mathematical relationship between the composition of the REIMS(or DESI)-accessible complex lipidome and the clinical behaviour of the disease. Due to the limited timeframe of the project, the lipidomic information was correlated with known histological markers having validated diagnostic and prognostic value, for instance receptor expression status or histological tumour type. The results of this correlation analysis clearly proved that practically all tested histological parameters (and thus the phenotype of the disease) can be determined using the complex lipidomic fingerprint of the tumour with sensitivities and specificities well exceeding 90 and 97 %, respectively.
summary for WP5
Several algorithms have been tested to identify marker fingerprints specific on tumor tissues. The marker screening activity will be continued as more data entries with appropriate histopatological classification will be generated along with the data collection. The tumour phenotyping task was performed using breast cancer and colorectal cancer and by the 19th month of the project the breast cancer results are available. Since breast cancer is one of the best characterized human malignancies, it served as an ideal model. The tumour phenotyping task focussed on establishing a mathematical relationship between the composition of the REIMS(or DESI)-accessible complex lipidome and the clinical behaviour of the disease. Due to the limited timeframe of the project, the lipidomic information was correlated with known histological markers having validated diagnostic and prognostic value, for instance receptor expression status or histological tumour type. The results of this correlation analysis clearly proved that practically all tested histological parameters (and thus the phenotype of the disease) can be determined using the complex lipidomic fingerprint of the tumour with sensitivities and specificities well exceeding 90 and 97 %, respectively.
(WP7) Impact and IP protection
Objective
As the aim of the project was to further the development to a stage where the technology already makes commercial sense. The goal was to inform and if possible to persuade potential future users and decision-makers about the advances of this novel technology.
Existing IP needed to be harmonised and prosecuted and if possible novel protectable IP created.
Outside experts needed to be informed about the technology and included in the project.
Dissemination and Workshops
The articles published, demonstrations and workshops held are summarised in the appropriate forms.
We trust the targeted audience has been reached. Scientific publications were directed at the scientific community in high impact factor papers, and we managed to reach the general public through appearances in the BBC news in the UK, Spiegel in Germany, a large number of local media in Hungary. Outside the participating countries, the project, amongst others, was covered by Chinese (China Daily), Middle eastern (AlJazeera) and US (WSJ, Bloomberg, Cnet). As a result a considerable interest was generated both among potential investors to the technology and potential cooperating partners from the medical and mass spectrometric community as well.
Follow up demonstrations were performed to both media and academic institutions widening the scope of the project both in terms of technology and geographical penetration.
Patent applications
In the project plan we set out to harmonise our patent applications and to protect any new developments that may be protected. During the entire project we were working together with Knobe a US based IP law partnership to streamline and reinforce our IP portfolio.
The Jet Desorption Ionisation (JEDI) application has been transferred to Knobbe from Danubia, the local patent office. It is only the Hungarian patent prosecution of this application that has remained in Hungary for cost effectiveness. The transfer was required for quality reasons as the local office was unable to handle the fine details of the application language to the required standard. The cost of this transfer was mainly financed outside of this grant.
A JEDI spin-off created in the US in the form of a divisional application has been further pursued. The concept is to create an umbrella protection above the other – more technology oriented – applications, at least in the US for the concept.
The REIMS patent application was granted in Mexico, and has – after a round of clarifications and amendments – positive feedback from the US patent office.
The REIMS spin-off for liquid samples has been pursued further as it provides extra protection for the technology enabling the REIMS based analysis of biological fluids. The application has entered national phase in a number of countries on 28.06.2014.
The last application we were keeping in our portfolio is the most technical and is trying to provide protection for the collision surface discussed in WP1. This is the application where we have been developing the most hardware related know-how during the project, but have not decided to try to turn it to protectable IP. This application has also entered national phase in a number of countries on 28.06.2014.
During the project the IP creation was mostly on refinements in already existing areas – such as the collision surface that was mentioned above – or in the development of algorithms and software code, that in itself does not constitute patentable subject matter. The created know-how is therefore mostly protected through the careful publication of the results and through internal project documentation.
Potential Impact:
The overall and ultimate purpose of our project was to enable the ISD device to become a diagnostic tool in the hands of practicing physicians, making their work easier, quicker and more effective. By reaching the market and becoming commercially applicable in the various fields of medicine, oncology, surgery, microbiology and gastroenterology, ISD technology could introduce a completely new paradigm into histopathology. That of objective information used at multiple locations simultaneously. This can result in harmonized diagnosis, prognosis and treatment methods leading to a more globalized medical community opening novel pathways.
The concept of the Intelligent Surgical Device was conceived to provide oncosurgery a tool that – without the need of modifying current surgical techniques – is capable of providing information to the surgeon that was previously not available in the surgical environment. By combining profiling mass spectrometry with the standard surgical toolset, we can provide information to the surgeon about tumor margin and recidive location. As a result the surgeon is able to make a more valued decision quicker, decreasing time spent in the operating room and potentially decreasing the need for repeated operations.
The target of the current Grant was to develop this technology to a status where it is accepted by the scientific community. The technology was developed to a status where under R&D conditions (the aim was specifically not to try to create a commercial clinical study ready prototype, but to prove) we could collect sufficient data to provide basis for the above claims. A Science: Translational Medicine article in June 2013 (“Intraoperative Tissue Identification Using Rapid Evaporative Ionization Mass Spectrometry” Julia Balog, …, Zoltan Takats) about the technology created huge public interest with demonstrations on the newly developing machinery to major newspapers and television channels worldwide. The publicity created resulted in a number of leads for potential cooperations from the scientific community as well as the medical community and potential industrial partners.
We started talks with a large number of medical centers worldwide for potential co-operations in data and experience collection with our devices, but due to our limited resources (multiple dedicated specifically modified mass spectrometers are required at each potential location) we are still concentrating on maintaining our capabilities within the consortium hospitals and in cooperating parties in Hungary.
We also received considerable interest from medical and analytical device manufacturers to collaborate for project execution. The most welcome of these was that of Micromass UK, a mass spectrometry company. Micromass/Waters devices were the 3rd type of devices we have been lately integrating to the project after using mass spectrometers from Thermo Scientific (US) and Tofwerk (Switzerland). Seeing the future development requirements in the project Micromass UK proposed that it is willing to finance the project further though the development and later in the clinical phase in return for ownership in the IP.
After a long negotiation phase it was acknowledged, that it is not possible to both continue the FP7 funding and use the resources of Micromass UK and have – in order to ensure the long term continuation of the project – opted to ask for an early termination of the current Grant. This, we believe ensures that the project receives the backing it requires in the long term future without further need for funding form the Taxpayers while keeping the project within the European Union. R&D activities are planned to be distributed between Hungary and the United Kingdom with a considerable increase in personnel. The number of data collection size is planned to be increased considerably in the next couple of years with more participating doctors and drawn in from an increased number of fields of oncology. The microbiology and food safety scope of the technology is to be increased dramatically in the next few years as well.
As a consequence we strongly believe that this SME startup led FP7 financed project has fulfilled its main targets – although some of the technical objectives have, admittedly, not yet been met – in that it developed the technology and the database to a level where its achievements supported its commercial viability to a level where industry funding was achievable.
List of Websites:
www.medimass.com
